WO2018197693A1 - Procédé et dispositif automatisés aptes à assurer l'invariance perceptive d'un évènement spatio-temporel dynamiquement en vue d'en extraire des représentations sémantiques unifiées - Google Patents
Procédé et dispositif automatisés aptes à assurer l'invariance perceptive d'un évènement spatio-temporel dynamiquement en vue d'en extraire des représentations sémantiques unifiées Download PDFInfo
- Publication number
- WO2018197693A1 WO2018197693A1 PCT/EP2018/060929 EP2018060929W WO2018197693A1 WO 2018197693 A1 WO2018197693 A1 WO 2018197693A1 EP 2018060929 W EP2018060929 W EP 2018060929W WO 2018197693 A1 WO2018197693 A1 WO 2018197693A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- signal
- unit
- value
- interest
- dynamic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
- G06V10/955—Hardware or software architectures specially adapted for image or video understanding using specific electronic processors
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/027—Frames
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/37—Determination of transform parameters for the alignment of images, i.e. image registration using transform domain methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/809—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of classification results, e.g. where the classifiers operate on the same input data
- G06V10/811—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of classification results, e.g. where the classifiers operate on the same input data the classifiers operating on different input data, e.g. multi-modal recognition
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
Definitions
- Automated method and device capable of ensuring the perceptive invariance of a spatio-temporal event dynamically with a view to extracting from it
- the present invention relates to an automated method and device capable of ensuring the perceptive invariance of a spatio-temporal event dynamically with a view to extracting unified semantic representations and memorizing them in association with a specific label. It also relates to a system for simplified memorization of the learning of networks of neuronal populations. The invention makes it possible to ensure a bijective relation between the space of the event, its contextual representation and its own representation.
- the invention is particularly applicable for the simplification of treatments of neuronal populations implemented on silicon for the analysis of various signals, including multidimensional signals such as the semantic representation of perception of images, sound or other modalities perceived independently. or simultaneously for example.
- signals including multidimensional signals such as the semantic representation of perception of images, sound or other modalities perceived independently. or simultaneously for example.
- the memory of the human being is fundamentally associative: we retain better when we can connect the new information to knowledge already acquired and firmly anchored in our memory. And this link will be all the more effective because it has a meaning for us.
- memory is today considered as a continual process of recategorization resulting from a continuous change of neuronal antagonistic pathways and the parallel processing of information in the brain.
- the perceptual invariance ensures a simplification of the categorization to be memorized by inscribing only one representation. The recall of an event, a scene or an object uses the inverse invariance function by putting the memorized information back in situation.
- the patent US2012308136 called "apparatus and methods forinstalle-code invariant object recognition” published on December 6, 2012 describes a bioinspired process based on action potential.
- the information is encoded in a variant as a pulse latency pattern with respect to an occurrence of a time event; For example, the appearance of a new visual frame or motion of the image.
- the pattern of the pulses advantageously is substantially insensitive to image parameters such as size, position and orientation, so that the identity of the image can be easily decoded.
- the associated device of the invention is based on the automatic perception of edges and curvatures described in the patent FR2858447, whose inventor is the same as the present application, "method and automated device of perception with determination and characterization of edges and of object boundaries, of a space, construction of contours and applications "and in which digital signals representing by parameters objects of a space are analyzed by a set of units or elementary modules for calculating histograms of same type, today bilinear and called "dynamic attractor".
- This device allows integration of the various processes into an electronic component.
- the invention thus relates to an automated method capable of calculating the perceptive invariance of an event occurring in a space, represented by pixels forming together a multidimensional x, y space, evolving over time at a succession of moments T, said data each associated with a temporal parameter A, B, ... being in the form of DATA (A), DATA (B), ... signals consisting of a sequence A xyt , B xyt , ... of binary number a bits associated with synchronization signals for defining the moment T of the space and the position x, y in this space, to which the signals A xy t, B xyt , ... received at a given instant, relative to at least one parameter for calculating the invariance of said event, according to the invention:
- a bilinear bilinear space-time histogram computation is performed, based on the elementary results of oriented edges and associated curvatures previously calculated, in order to identify an area of interest of the space, related to the event, according to the statistical criteria of curvature and oriented edge applied to a spatial parameter, in order to recover, for the area of interest thus identified at the end of the moment T: at. a median value of edge oriented bo 0 and curvature cb 0 ,
- a zone of interest is selected in the upstream repository x, y, before scaling k, by giving it a repository X, Y, such that these new coordinates are defined according to the old coordinates x, y by a translation of the origin ki 0 , kj 0 and a rotation of the reference frame by an angle b 0 , such that:
- the scale of this area of interest is adapted so that it fits in a square centered by 2 Z pixels, truncating the number of bit u constituting the range k.ap 0 or k. bp 0 the largest, in order to keep only the z most significant bits of this range.
- the scaling of factor k is parametrically carried out by a Gaussian filtering calculation of size 2w + 1, the value w being the parameter of the filtering matrix, in association with a factor decimation function k on the space x, y for a moment T,
- a calculation of the parameters w and k is performed as a function of the median value of curvature cb 0 defined in the previous step d), by comparing the new median curvature value cb 0 with two bounds of increasing values La and Lb, for 0 ⁇ w ⁇ w_max, such as: at. cbo ⁇ The value of w is increased by one.
- Flag Val 0 b. The ⁇ cbo ⁇ Lb the value of w remains unchanged.
- Flag Val 1
- iteration is carried out by iteration, at the end of each moment T, of an update of Gaussian filtering parameters w and of decimation k in step a) in order to maintain the value of median curvature cb 0 between two increasing bounds.
- the scaling of factor k is parameterized by the use of a dedicated sensor on the space x, y during a time T,
- a zone of interest is selected in the upstream reference system x, y, by the use of a dedicated sensor on the space x, y during a time T, by assigning it a zone dimension k.ap 0 by k.bp 0 in an X, Y repository of this space x, y, such that these new coordinates are defined according to the old coordinates x, y, by a translation of the origin ki 0 , kj 0 a rotation of the reference frame an angle bo 0 , and a ratio of homothety 1/2 (u "z) ,
- each of these data is transformed into a global, dynamic and structural type positioned in the repository X1, Y1,
- a bilinear bilinear space-time histogram computation is carried out on the basis of the elementary results of global, dynamic and structural type positioned in the repository X1, Y1 calculated previously, in order to identify an integrated area of interest. in a square of 2-dimensional Z side of the space, representing part of the event, according to global, dynamic and structural statistical criteria to a spatial parameter repository X1, Y1, in order to recover for the zone of interest thus identified at the end of the moment T, at least, a structural semantic representation:
- a new bilinear space-time histogram computation is reiterated, by inhibiting the information of the previously identified main zones so as to identify other areas of interest within an area of the space 2 Z square of uninhibited side, up to n computation sequences or when a remaining uninhibited zone of this space no longer produces an area corresponding to said statistical criterion, in order to recover it, for the zone of interest thus identified at the end of the moment T, at least, a structural semantic representation:
- the invention also relates to a device for calculating the automatic perceptual invariance of an event occurring in a space, represented by pixels forming together a multidimensional x, y space, evolving over time at a succession of moments T, said associated data.
- a time parameter A, B ... being in the form of DATA (A), DATA (B), ... signals consisting of a sequence Axyt, Bxyt, ... of a bit number of a bits associated with synchronization signals for defining the moment T of the space and the position x, y in this space, to which the signals Axyt, Bxyt, ... received at a given instant, with respect to at least one parameter for the invariance calculation of said event.
- the device for calculating automatic perceptual invariance comprises:
- a looped bilinear space-time histogram computation unit based on the elementary results of oriented edges and associated curvatures previously calculated, in order to identify an area of interest of the space, related to the event , according to the statistical criteria of curvature and oriented edge applied to a spatial parameter, in order to recover it, for the zone of interest thus identified at the end of the moment T:
- a servo unit implemented if necessary, by iteration at the end of each moment T, by updating the change of scale parameters k in step a) in order to maintain the median curvature value cb 0 between two increasing limits.
- the scaling factor calculation unit k consisting of a Gaussian filtering calculation of size 2w + 1, the value w being the parameter of the filtering matrix, in association with an affine function of the factor k decimation parameter w on the space x, y for a moment T, It comprises, the unit of calculation of the parameters w and k is a function of the median value of curvature cbO defined in the previous step d), by comparing the new value of median curvature cb0 to two bounds of increasing values La and Lb, for 0 ⁇ w ⁇ w_max, such as:
- the servocontrol unit by iteration at the end of each moment T, update gaussian filtering w and decimation parameters k in step a) in order to maintain the median curvature value cb 0 between two increasing limits.
- the zone of interest selection unit in the upstream repository x, y is made by the use of a dedicated sensor on the space x, y during a frame sequence, by assigning it a dimension of zone k.ap 0 by k.bp 0 in a reference X, Y, of this space x, y, such that these new coordinates are defined according to the old coordinates x, y, by a translation of the origin ki 0 , kj 0 a reference rotation of an angle bo 0 , and a homothety ratio 1/2 (u "z) ,
- a translation unit transforming each of these data into a global, dynamic and structural type positioned in the repository X1, Y1, b) a bilinear bilinear space-time histogram computation unit, calculating, from the basic results of type global, dynamic and structural positioned in the repository X1, Y1 previously calculated, in order to identify an area of interest integrated in a square of dimension Z 2 of space side, representing a part of the event, in global, dynamic and structural statistical criteria to a spatial parameter of repository X1, Y1, in order to recover, for the area of interest thus identified at the end of time T, at least, a structural semantic representation:
- the device comprises means for associative storage between the semantic representation ensuring the invariance and its associated label, the label representing the event in a unified manner, and the representation of the area of interest identified, its context.
- FIG. 1 is a representation, in its generality, of the automated perceptive invariance method allowing the invariant extraction of the perceived semantic representations of the event in its context according to the invention
- FIG. 2 is a representation, in its generality, of the perceptual invariance device according to the invention.
- FIG. 3 is a detailed description of the process of calculating parameters allowing perceptive invariance by the use of a dynamic attractor;
- FIG. 4 is an application integrating the visual perceptual invariance by opto-mechanical implementation of two visual sensors;
- FIG. 5 is a use of a vision device integrating two units of semantic, context and local representation of a label (100), in relation with an associative memory (10);
- FIG. 6 is an example of a semantic representation unit of an associative memory label (100) in combination of dynamic attractor units (80_i) associated with a translation transfer unit (71);
- Figs. 7a and 7b are illustrations of the transfer function of the translation unit (71);
- FIGS. 8a to 8d are illustrations of the organization of the bilinear histogram calculations of the dynamic attractor unit (80_i);
- FIG. 9 is an illustration of the extraction, by a dynamic attractor (80_0), of contextual semantic representations of a perceived element of the vision device of FIG. 5.
- FIG. 1 describes the generic process of perceptive invariance from an example of dynamic management, in sequences, of the interaction between the event, signal (HD) referenced (x, y), its context, signal (MAP1 ) referenced (i, j), and its representation, signal (MAP2) referenced (X, Y).
- MAP1 two-dimensional data
- a first moment T starts at time (t0) by the representation of the event in its context which must be global, like a homogeneous convex region of general curvature (cb 0 ) sufficient.
- the filter parameter w equal to zero defines the output stream (MAP1) which is analyzed a computation of curled bilinear histograms corresponding to a dynamic attractor (80_0) and deduces therefrom at the end of the moment T a curvature value (cb 0 ) .
- This curvature value (cb 0 ) is compared with two bounds of increasing values La and Lb, for 0 ⁇ w ⁇ w_max:
- the associated values of w and k are updated for the next T time.
- the sequence of moments T (t3) to (t5) involve dynamic attractors by dynamic recruitment in order to recover, for the zone of interest thus identified at the end of the moment T , a structural semantic representation, by inhibiting the information of the main zones previously identified so as to identify other areas of interest within a zone of the square space of 2 Z of uninhibited side.
- the first dynamic attractor (80_1) in moment T (t3) focuses on the most important accumulation of points, here the curvature cbi and its orientation boi of the figure "2" its position ⁇ - ⁇ , ⁇ and its dimension ap-i, bp-i, then, by inhibiting the information of the previously identified main zone, at time T (t4), a second dynamic attractor (80_2) extracts its horizontal base, curvature cb 2 and its orientation bo 2 , its position x 2 , y 2 and its dimension ap 2 , bp 2 , the first dynamic attractor remains hooked and finally, at time T (t5), a third dynamic attractor (80_3 ) extracts its oblique curvature cb 3 and its orientation bo 3 , its position x 3 , y 3 and its dimension ap 3 , bp 3 .
- FIG. 2 represents the perceptual invariance device in association with the method described above in FIG. 1, and comprises the following units:
- the transduction unit (130) The transduction unit (130)
- An event (90) is perceived by a transducer (130) which outputs a data stream (HD xyt ), associated with pixels together forming a multidimensional space evolving over time and represented at a succession of moments T, said data associated with each to a time parameter HD, being in the form of DATA (HD) signals, consisting of a sequence (HD xyt ), a binary number of a bits synchronized by a clock (Ck) and the x, y positions of dimension respectively of o and p bits, of a data sequencer (150).
- This stream (HDx y t) is connected to the input of a Gaussian filter unit (200) and to the input of a buffer memory (600).
- the decimation unit (Dec) is composed of:
- a data sequencing unit (150) generating the x and y addresses as well as data synchronization signals in this decimation unit (Dec),
- a gaussian filtering unit (200) synchronized by a clock (Ck) and the x, y positions of a data sequencer (150) and parameterized by the filtering value (w).
- This unit performs a Gaussian-type two-dimensional filtering function by convolving a Gaussian matrix of size 2w + 1 on the input data, the result is a two-dimensional data stream lmG ( Xy t), output in sequence to a unit of decimation (300).
- a decimation unit (300) receiving the input stream lmG (Xyt) and outputting the two-dimensional data stream ⁇ ⁇ to the oriented edge calculation unit (350), selecting in two dimensions x and y, an element all the k elements.
- the conversion unit (71 ') is composed of:
- An oriented edge calculation unit (350) transforming each bit-coded incoming data item i, j, of the flow MAP1 (i jt) into an angular value Bo, position i, j, corresponding to the presence of 'an edge indicated by the flag (Valid) to one or a neutral value if not, flag (Valid) to zero.
- These values constitute the stream ImBo (ijt) coded on a bit
- a bending calculation unit (360), receiving a flow ImBo (ijt), in order to calculate a notion of curvature (Cb), position i, j, corresponding to the angular local variations found or to a neutral value otherwise, flag (Valid) to zero.
- These values constitute the stream ImCb (ijt) coded on a bit,
- the dynamic attractor (80_0), described more precisely in FIG. 3, calculates two-dimensional spatio-temporal histograms at each elementary moment validated by the flag (Valid) to one on the instantaneous truncated data pairs i, j, and Bo, Cb each pair comprising 2z bits as a set allowing calculation of two-dimensional spatio-temporal histograms delivering at the end of the moment T a result, if presence of the element, A center of gravity position of the element i 0 , jo, and a zone binary signal Z_ROI
- the unit of calculation of the perceptive invariance (Inv) comprises:
- the associated values of w and k are transmitted for the following time T, (w) at the unit (200) and (k) at the unit (300).
- a selection unit of the region of interest receives the coordinates of the center of gravity i 0 , jo, the coefficient k of decimation, the angular orientation bo 0 , and the coordinates x, y of the stream (HD (xyt) ), and calculates new coordinates X0, Y0 according to the function:
- An address multiplexer (520) transfers the previously computed addresses, for the valid area of interest (Z_ROI), to the address port of the buffer (600).
- Wr write command
- the ROI unit comprises:
- a scaling unit (542) truncating the X1 and Y1 values on the most significant z bits to generate an X and Y address each over a range of 2 Z values, corresponding to the size of the square of 2 Z aside.
- MAP2 The data flow (MAP2) resulting from the reading of the memory (600) in address X, Y, corresponding to the invariant information of the perceived element.
- FIG. 3a describes the process of calculating the parameters allowing the perceptual invariance by the use of a dynamic attractor (80_0) consisting of two sets of bilinear histogram calculation units (81_S) and (81_P) receiving the signals and each producing a classification value whose product (V) serves as validation of the calculations:
- the first subset (81 _S) receiving two signals carrying the time parameters Bo, Cb and,
- the two subsets jointly producing a binary signal Z_ROI representing a zone of interest and a binary signal BC representing the value of the temporal parameter of this zone,
- An AND operator 86 combining the output binary signals Z_ROI and BC, the output of this operator serving as validation signal (78) for calculating sets of bilinear histogram calculation units,
- An external sequencer initializing the memories of these sets of bilinear histogram calculation units at the start of the moment T, then validating the calculation during the moment T, to identify an area of interest of the space linked to the event , according to the statistical criteria of curvature and oriented edge applied to a spatial parameter and at the end of the moment T, recovering for the area of interest thus identified:
- Each set of bilinear histogram calculation unit (81_S) and (81_P) comprises:
- An analysis memory unit (82) having memories with addresses, each associated with possible values of 2.2 Z bits of the signal Bo, Cb for the unit (81_S) and i, j for the unit ( 81_P) and whose writing is controlled by a "Write" signal
- the classification unit (83) operates in the same way by replacing the parameters Bo, Cb, by i, j.
- the classification binary signal is validated if the first parameter is between criteria A and B, and at the same time if the second parameter is between criteria C and D.
- Figure 3b shows the structural result of the element calculated by the spatio-temporal bilinear histograms during the moment T.
- the pair of parameters Bo and Cb form a 2D representation, with the origin of the upper left corner of a horizontal square of dimension 2 Z -1, Bo on one axis and Cb on the other axis.
- a third vertical axis represents the accumulation of each couple.
- the cumulation maximum is represented by the pair of values bo 0 , cb 0 .
- FIG. 3c represents the spatial result of the element calculated by the spatio-temporal bilinear histograms during the moment T.
- the pair of parameters i and j form a 2D representation, with as origin the upper left corner of a square horizontal dimension 2 Z -1, i on one axis and j on the other axis.
- a third vertical axis represents the accumulation of each couple.
- the maximum cumulation is represented by the pair of values i 0 , j0.
- the spatial range of the element has for dimension apO and bpO centered on i 0 , jo, and oriented by an angle equal to bo 0 centered in i 0
- Figure 3d shows the region of interest ROI, spatial result of the calculated element previously reported on the upstream repository x, y.
- the data of this zone of interest ROI are defined according to the coordinates X and Y, origin in the center of the zone of interest and dimension -k.ap 0 ⁇ X ⁇ k.ap 0 and -k.bp 0 ⁇ Y ⁇ k.bp 0 .
- FIG. 4 describes a generic device for perceptive invariance of an event (90) from two coupled cameras, the first camera C1 outputting a low resolution MAP1 signal associated with pixels together forming a multidimensional space evolving over time and represented at a succession of moments T, on a large field and the second camera C2 movable relative to the first high resolution on a narrow field issuing a MAP2 signal associated with pixels together forming a multidimensional space evolving in time and represented in FIG. a succession of moments T.
- the event perceived by the camera C1 is analyzed globally by the semantic representation unit of a label (100_1) which provides information,
- FIG. 5 depicts a generic perceptual invariance device from a single, high resolution, wide field camera outputting an HD signal. It is an extension and generalization of the device described in FIG.
- the unit of generic semantic representation (100_0), described in detail in FIG. 6, receives, as input, the signal ⁇ ⁇ , its position (i, j) and incorporates an associative memory unit (10) in association with:
- a translation transfer unit (71) which includes the unit (71 ') and is extended to the global and dynamic perception, described in detail in FIG. 7a and,
- a dynamic attractor unit (80_0) that includes, as described in FIG. 2,
- This unit of generic semantic representation (100_0) delivers, at the end of the moment T, the information to the invariance calculation unit as previously described in FIG. 2 as well as to an associative memory unit by adding to it additional information of elementary semantic representations global and dynamic, the set of these representations forms a message that is read by the associative memory (10) and delivers a label value (Loutj). Conversely, this label value can be reintroduced into this same associative memory (10) as a value Lin i, the output of this memory then provides a message, which has been previously learned, which drives the classifiers (84) for a confirmation of information acquired.
- the generic semantic representation unit (100_1) receives, as input, the signal MAP2 ( xYt), its position (X, Y) and having n dynamic attractor units, delivers up to n sub-messages to an associative memory (10). ) which deduces, if already known, a label (LoutJ). As previously described, an inverse function, which introduces the label (LinJ) into the associative memory (10) of the unit (100_2), drives the classifiers (84) to confirm the acquired information.
- FIG. 6 illustrates an example of use of the associative memory (10) in combination of dynamic attracting units (80_i) and of a translation transfer unit (71) defining the semantic representation unit of a label ( 100). For clarity, the sequencing signals have been omitted.
- Spatiotemporal data (70) from an upstream element, receiver (92) or associative memory unit processing results (10), not shown here, are delivered on the input port (E, P (i, j)) of a translation transfer unit (71) which in turn synchronously outputs, clocked by a clock signal (Ck), elementary semantic representations referenced in position on its output ports (G ), (D), (S), and (P).
- Each output port (G), (D), (S), and (P) is connected independently and respectively on the G bus (72), the D bus (73), the S bus (74) and the bus.
- P (75) all of identical size of 2z bits.
- the n dynamic attractor units (80_1) to (80_n) connect to these four buses, respectively on their input port (G), (D), (S), and (P).
- the dynamic attractor unit (80_1) The dynamic attractor unit (80_1)
- the dynamic attractor unit (80_1) is shown in greater detail in order to explain its operation, knowing that all the dynamic attracting units (80_1) to (80_n) are of identical constitution.
- This dynamic attractor unit (80_1) includes:
- Each statistical processing unit (81_x) comprises:
- a bilinear histogram calculation unit (82), comprising ⁇ A given input (x) corresponding to (G) or (D) or (S) or (P) according to the statistical processing unit (81 _x)
- ⁇ a sequencing unit, not shown here, depending on the operating mode, either sequentially or by number of events, which ensures cyclically in sequence, the initialization phase, the histogram calculation stage, the phase update registers (R) and the automatic classification phase.
- the initialization phase consists of zeroing the storage memory of the histogram calculations and initializing the various calculation registers.
- each data item (x) presented corresponds to an input signal (V) that validates the calculation or not.
- the registers (R) are updated as well as the registers of the automatic classification unit (83).
- the calculated values are, the number of calculations (NBPTS), the median (Med), the value of the maximum (RMAX), its position (PosRMX) and the classification limits (A), (B), (C), and (D).
- o two classification units, automatic (83) and request (84) each receive the coded data on 2z bits of the input port (x) and each deliver a valid classification binary signal for the z bits of high weight between its classification terminals (A) and (B) and for the low-order z bits between its classification terminals (C) and (D),
- a Boolean classification validation unit (85) receives the binary classification signals of the two automatic classification units (83) and request (84). The logical AND between these two binary classification signals is transmitted out of the statistical processing unit (81 _x).
- a Boolean space-time classification unit (86) receives the binary classification signals from the four statistical processing units (81_G), (81_D), (81_S), and (81_P) to form a logical AND which is transmitted to the histogram calculation validation unit (87).
- a histogram calculation enable unit (87) comprises a two-input AND logic unit, one inverted (88) and a two-input OR logical unit (89).
- the logical unit AND (88) receives in live the binary signal coming from the logical unit AND (86) and reverses the input binary signal (Cin) of the unit (80_1) and delivers a binary calculation validation signal of histogram on the input (V) of each statistical processing unit (81_G), (81_D), (81 _S), and (81_P).
- the OR logic unit (89) receives the input binary signal (Cin) from the unit (80_1) and the histogram calculation enable binary signal of the AND logic unit (88) and outputs a binary signal of Inhibit on the output port (Cout) of the unit (80_1).
- An output register unit (76) has the registers (RSi-1) to (RSi-q) updated each time the value (NBPTS) is exceeded relative to an externally parameterized threshold.
- the order of the registers (RSi-1) to (RSi-p) corresponds to the median values (Med- ⁇ , Med 2 ) and to the classification range (Pi, P 2 ) defined by the difference between the terminals of classification (B) minus (A) and (D) minus (C) for each statistical processing unit (81_G), (81_D), (81 _S), and (81_P).
- the output register unit (76) includes the registers (RSi-1) to (RSi-8).
- a number of registers (RSi-x) are not used because they are irrelevant.
- the visual perception of a text has a uniform overall mode (same colors and non-displacement), only the structuring aspect brings relevant information, which reduces the eight starting registers to three: centroid, dimension and structure.
- An input register unit (77) has the registers (RSo-1) to (RSo-q) representing the same organization as that of the output register unit (76).
- the Associative Memory unit (10), described in the preceding figures, in its generic implementation has for interface with the dynamic attracting units (80_1) to (80_n) the message (MEin_i) consisting of n sub-messages (RSin_1) to (RSin_n) and the message (MEoutJ) consisting of n sub-messages (RSout_1) to (RSout_n).
- Sub-message (RSin_1) is transmitted from the output register unit (76) of the dynamic attractor unit (80_1) to the input port (In) of the memory sub-unit (2_1) of the unit Associative memory (10).
- the sub-message (RSin_2) is transmitted from the output register unit (76) of the dynamic attractor unit (80_2) to the input port (In) of the memory sub-unit (1_2) of the associative memory unit (10), and the transmission continues in the same order up to the rank n.
- the sub-message (RSout_1) is transmitted from the output port (Out) of the memory sub-unit (2_1) of the associative memory unit (10) to the input register unit (77) of the dynamic attractor unit (80_1).
- the sub-message (RSout_2) is transmitted from the output port (Out) of the memory sub-unit (1_2) of the associative memory unit (10) to the input register unit (77) of the dynamic attractor unit (80_2), and the transmission continues in the same order up to the rank n.
- the associative memory unit (10) comprises:
- a first set composed of n memory subunits, each composed of 2 m words of m bits, denoted by (2_1) to (2_n), and each receiving on their input port (In) respectively the sub-message ( RSin_1) for the memory sub-unit (2_1) to the sub-message (RSin_n) for the memory sub-unit (2_n),
- a second set consisting of a memory sub-unit of 2 V words of v bits (1) receiving on the input port (In) the label (LinJ) and,
- a maximum likelihood calculation unit (4) to select the most represented r, s or other value. This unit (4) receives from the output port
- each memory subunit (2_1) to (2_n) a value r, or others respectively on an input port (L_i) to (L_n) with their respective validating signal on the input respectively (V_1) at (V_n).
- the internal sequencing is provided by a clock signal (CK) introduced into the unit (4).
- CK clock signal
- the choice of the maximum likelihood is set on the output port (L_i), a v-bit bus passes this value to the input port (Adr) of the memory sub-unit (1) which delivers on its output port (Out) the value of the label (Lout_i).
- the presentation of (LinJ) on the input port (In) of the memory sub-unit (1) causes the delivery of the value j on its output port (Cadr), this value j is transmitted to the bus (AB) through a link value choice unit (10) and is presented to all memory subunits (2_1) to (2_n) which each delivers on its output port (Out) the -message respectively (RSout_1) to (RSout_n) which together form the message (MEoutJ).
- the sub-messages (RSin_1) to (RSin_n), corresponding to the message (MEin_i), are respectively presented on the input port (In) of each sub-memory unit (2_1) to (2_n ) which each deliver a value r, or others on their respective output port (Cadr) in association with a validating bit signal outputted by the output (M) of the same memory subunit.
- the sub-message presented is absent from the memory sub-unit, it delivers on its output (M) a binary signal of non-validation, the value present on its output port (Cadr) is then ignored.
- Each received message (MEin) of n.m bits is composed of n sub-messages (RSin_x) of m bits, x varying from 1 to n.
- each message (MEout) of n.m bits delivered by the associative memory is composed of n sub-messages (RSout_x) of m bits, x varying from 1 to n.
- Each sub-message is segmented into q elements (RSi_x) entering or (RSo_x) outgoing z bits corresponding to m / q bits whose rank of the element corresponds to a notion of position, sizing and characterization.
- the position is defined by its reference (Ref), generally varying from one to three, often equal to two for a pair of elements, for example x and y representing a relation between two distances in the reference frame (Ref), or t and f representing a relationship between time and frequency in the reference frame (Ref).
- Ref reference
- x and y representing a relation between two distances in the reference frame (Ref)
- t and f representing a relationship between time and frequency in the reference frame (Ref).
- it is the position of the barycenter of the data cloud representing the characterization defined above by its sub-message elements.
- the dimensioning characterizes the extent of the data cloud, generally its dimension, therefore an element (RSi_x), in each of the axes of the reference frame (Ref).
- Characterization is usually an elementary semantic representation of type:
- ⁇ Dynamics for example non-limiting, a movement is defined by its speed and its orientation, it is also the prosody of a voice, etc.
- an edge is defined by its orientation and curvature
- a phoneme is defined by the distribution of its formants in time, etc.
- the label consists of a word of v bits, the amount of recordable labels is 2 V - 1, the label "zero" being excluded.
- FIG. 7-a explains the operation of the translation transfer unit (71) from the spatio-temporal data (70) (time data E and position P (i, j)) originating from an external sensor, represent.
- Each spatio-temporal data (70), coming into this unit (71), is translated and delivered on four output ports, in a synchronous manner by a signal (Ck), into three distinct elementary semantic representations, (G), (D), (S), positioned at (P).
- Each output port (G), (D), (S), and (P) is connected independently and respectively on the G bus (72), the D bus (73), the S bus (74) and the bus. P (75).
- Figure 7-b is a pictorial representation showing the registration of the different data (G), (D), (S), and (P).
- the incoming data is represented in its global mode output (G), dynamic output (D), and output structural (S), and position (i, j) determined by the data (P), according to three plans recorded in 2D mode .
- the position (P) is expressed according to the size of its base.
- 2D for visual (x, y) or auditory (t, f) data it can be extended to 3D or reduced to 1D, higher dimensions are possible but unacceptable by the devolved memory size.
- FIG. 8 illustrates the organization of the results of the calculations of the four bilinear histograms of the dynamic attractor unit (80_i) from the data (G), (D), (S), and (P) 2z bits from the translation transfer unit (71).
- the incoming data processed in this example is 2D vision type.
- the unit (71) translates this data into
- Each incoming data is coded in a 2z-bit word giving a 2 Z x 2 Z matrix representation of the histogram calculation, the first z bits representing an axis and the z bits remaining the second axis of the matrix.
- an edge portion of an object (Ob) is visualized in FIG. 8d by its representation in position (P), gray values corresponding to the classified results, by the classification unit (83). ), the calculation of bilinear histogram of the matrix (H_P). The result of this histogram calculation is transmitted to the output register unit (76) with the value of its barycenter at 2z bit position (x, y) and its 2z bit range (ap, bp). .
- the orientation and the perceived local curvature of the object (Ob), FIG. 8c, is delivered by the calculation of the bilinear histogram (H_S) whose result of the calculation is transmitted to the output register unit (76). with its value, its center of gravity, so its semantic representation of orientation and curvature of 2z bits (bo, cb) and its tolerance of 2z bits (as, bs).
- H_S bilinear histogram
- FIG. 8a indicates, by the result of the bilinear histogram calculation (H_G) the dominant color of the part of the object (Ob) represented by its hue and saturation value of 2z bits (t, s) with its value. 2z bit tolerance (ag, bg), transmitted to the output register unit (76).
- H_G bilinear histogram calculation
- FIG. 8b shows, by the result of the bilinear histogram calculation (H_D), the local displacement of the part of the object (Ob) represented by its movement direction value and its speed on 2z bits (dir, vit ) with its tolerance value of 2z bits (ad, bd), transmitted to the output register unit (76).
- the input register unit (77) updates, in the same order, the classification terminals of the query classification units (84) of each statistical processing unit (81 _G), (81_D), ( 81 (S), and (81 (P).
- This perceptual process enslaves data perceived, represented and interpreted as a learned label.
- the incoming sub-message (RSin_x) of the associative memory (10) is composed of the results (t, s, ag, bg) for the sub-message (RSi-1) and (RSi-2) , (dir, vit, ad, bd) for the sub-message (RSi-3) and (RSi-4), (bo, cb, as, bs) for the sub-message (RSi-5) and (RSi-6) , and (x, y, ap, bp) for the sub-message (RSi-7) and (RSi-8).
- RSout_x the outgoing sub-message (RSout_x) of the Associative Memory (10).
- This sub-message (RSin_x) is a basic semantic representation global, dynamic, structural (answer to the question What?) And positioned (answer to the question Ou?).
- the n sub-messages (RSin_x), x varying from 1 to n, define the message MEin_i representing at the output of the associative memory (10) the label (Loutj).
- FIG. 5 represents an application using visual perceptual invariance with extraction of elementary semantic representations recorded in an associative memory (10).
- HD high density video signal
- This signal (HD) is delivered, together with:
- a decimation unit (Dec) which transforms by spatial decimation this signal (HD) into a low resolution signal (MAP1) representing the same point of view, for example a VGA format at 0.3 Mp and with the same frame rate fps, and, • a region of interest extraction unit (ROI) which splits this signal (HD) into a signal (MAP2) of the same spatial resolution over a range corresponding to and transmitted at the same frame rate 50 fps or at a multiple of the frame rate, for example 8 times.
- MAP1 low resolution signal
- ROI region of interest extraction unit
- the video signal (MAP1) is introduced into a generic semantic representation unit (100_1) which perceives the written symbol as a task.
- the attractor dynamic (80_0) of this unit (1 00_1), see figure 9, delivers the elementary semantic representations:
- G_1 global (G_1) words of 2z bits transformed into elementary semantic representation of color; barycentre hue and saturation (t 0 , s 0 ) and range of distribution of data along the two axes T and S (ag 0 , bg 0 ) respectively corresponding to sub-messages (RSin-1) and (RSin-2).
- D_1 dynamic (D_1) words of 2z bits transformed into elementary semantic representation of movement; barycentre direction and speed of movement (dir 0 , vit 0 ) and range of distribution of the data along the two axes Dir and Vit (ad 0 , bd 0 ) respectively corresponding to sub-messages (RSin-3) and (RSin-4) ).
- the set of elements (RSin_1) to (RSin-8) constitutes the sub-message (RSin_0).
- the invariance calculation unit (Inv) reads through the communication bus (S_MA), the oriented edge and curvature information (bo 0 , cb 0 ) and calculates the filter coefficient w and decimation k which is transmitted to the unit (Dec).
- This loop dynamically adapts the decimation of the signal (HD) to (MAP1) in order to maintain the result cb 0 between the two terminals La and Lb, sets the flag Val to one and, triggers the calculation of the invariant flow (MAP2), dimension 2 z x2 z , equal to:
- the sequence of the moments T involve dynamic attractors by dynamic recruitment in order to recover, for the zone of interest (ROI) thus identified at the end of the moment T, a semantic representation structural, by inhibiting the information of the main zones previously identified so as to identify other areas of interest within a zone of the square space of 2 Z of uninhibited side.
- This image zone referenced in position ( ⁇ 2 ( ⁇ ⁇ ) ) forms the video signal (MAP2) which is introduced into a generic semantic representation unit (100_2) in order to perceive the written symbol in its integrity.
- the curvature of digit two has the largest number of perceived pixels and is therefore represented by the dynamic attractor (80_1) which delivers an m-bit sub-message (RSin_1) composed of oriented edge elements and associated curvature.
- RSin_1 m-bit sub-message
- the dynamic recruitment of a second dynamic attractor (80_2) which receives the inhibition of the preceding processing, perceives the largest number of pixels corresponding to the horizontal part of the number two and delivers a sub-message (RSin_2) of m bits consisting of 2z bits oriented edge elements and associated curvature (bo 2 , cb 2 ), 2z bits position elements (x 2 , y 2 ), 2z bits dimension (ap 2 , bp 2 ) and, orientation cc 2 equal to bo 2 .
- RSin_2 sub-message
- the sequence continues with the dynamic recruitment of a third dynamic attractor (80_3), which receives the inhibition of the previous treatments, perceives the largest number of pixels corresponding to the oblique part of the two digit and delivers a sub-message ( RSin_3) of m bits consisting of 2z bits oriented edge elements and associated curvature (bo 3 , cb 3 ), 2z bit position elements (x 3 , y 3 ), of dimension (ap 3 , bp 3 ) of 2z bits and, cc 3 orientation equal to bo 3 .
- the number of pixels remaining untreated being less than a qualification threshold (value (NBPTS) below a threshold), the recruitment sequence of a new dynamic attractor stops.
- NPTS qualification threshold
- the message (MEin_i) consists of sub-messages (RSin_1), (RSin_2), and (RSin_3), a combination of 3 words of 6z bits.
- RSin_1 sub-messages
- RSin_2 sub-messages
- RSin_3 sub-messages
- 2 6z an input value given z coded on 4 bits, 2 6z is equal to 2 24, that is to say nearly 1 6 million values
- This message (MEin_i) is associated with the label (Lin i) worth “two", in this case, and is stored in the associative memory (10) of the semantic representation unit (100_2).
- the sequencing unit (Sec) controls, via the communication buses (S_P2) between the semantic representation unit of a label (100_2) and the sequencer (Sec) and (S_MA) between the associative memory unit ( 10) and the sequencer (Sec), the organization of perceived messages.
- An associative memory (10) associates the label (LoutJ) from the unit (100_1) and the label (Loutj) from the unit (100_2) in order to output a label (Lout_k) corresponding to the value of the number "two In context. For example, this device validates the "one" digit, contextually positioned to the right of the "two” previously perceived to form the number "twenty-one".
- the description of the device for calculating automatic perceptive invariance of an event described in FIG. 5 can advantageously be integrated into an electronic module and be used as intelligent control of an effector (97). ) from signals from a transducer (130), in the field of the Internet of Things.
- MEoutJ outgoing message grouping n sub-messages (RSout_1) to (RSout_n) n number of sub-messages entered (RSin_i) or output (RSout_i)
- V_i input validation value (L_i)
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Description
Claims
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020197028748A KR102454841B1 (ko) | 2017-04-28 | 2018-04-27 | 통합된 의미론적 표현들을 추출하기 위하여 공간-시간 이벤트의 다이나믹 지각 불변량을 제공가능한 자동화된 방법 및 디바이스 |
| US16/499,820 US11164049B2 (en) | 2017-04-28 | 2018-04-27 | Automated method and device capable of providing dynamic perceptive invariance of a space-time event with a view to extracting unified semantic representations therefrom |
| EP18718863.6A EP3616132A1 (fr) | 2017-04-28 | 2018-04-27 | Procédé et dispositif automatisés aptes à assurer l'invariance perceptive d'un évènement spatio-temporel dynamiquement en vue d'en extraire des représentations sémantiques unifiées |
| CN201880025028.7A CN110520869B (zh) | 2017-04-28 | 2018-04-27 | 信号处理电路和方法、从信号提取感知不变的装置和方法 |
| JP2019554641A JP6909307B2 (ja) | 2017-04-28 | 2018-04-27 | 一体化された意味論的表現を時空事象から抽出することを目的として時空事象の動的知覚的不変性を提供することができる自動化方法及び装置 |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| FR1700465 | 2017-04-28 | ||
| FR1700465A FR3065825B1 (fr) | 2017-04-28 | 2017-04-28 | Procede et dispositif automatises aptes a assurer l'invariance perceptive d'un evenement spatio-temporel dynamiquement en vue d'en extraire des representations semantiques unifiees |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2018197693A1 true WO2018197693A1 (fr) | 2018-11-01 |
Family
ID=59859122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/EP2018/060929 Ceased WO2018197693A1 (fr) | 2017-04-28 | 2018-04-27 | Procédé et dispositif automatisés aptes à assurer l'invariance perceptive d'un évènement spatio-temporel dynamiquement en vue d'en extraire des représentations sémantiques unifiées |
Country Status (7)
| Country | Link |
|---|---|
| US (1) | US11164049B2 (fr) |
| EP (1) | EP3616132A1 (fr) |
| JP (1) | JP6909307B2 (fr) |
| KR (1) | KR102454841B1 (fr) |
| CN (1) | CN110520869B (fr) |
| FR (1) | FR3065825B1 (fr) |
| WO (1) | WO2018197693A1 (fr) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11536805B2 (en) | 2018-06-25 | 2022-12-27 | Silc Technologies, Inc. | Optical switching for tuning direction of LIDAR output signals |
| US12429569B2 (en) | 2019-05-17 | 2025-09-30 | Silc Technologies, Inc. | Identification of materials illuminated by LIDAR systems |
| US11650317B2 (en) | 2019-06-28 | 2023-05-16 | Silc Technologies, Inc. | Use of frequency offsets in generation of LIDAR data |
| US11027743B1 (en) * | 2020-03-31 | 2021-06-08 | Secondmind Limited | Efficient computational inference using gaussian processes |
| CN113886866B (zh) * | 2021-08-09 | 2024-06-07 | 安徽师范大学 | 基于语义位置转移的时空关联轨迹隐私保护方法 |
| US12411213B2 (en) | 2021-10-11 | 2025-09-09 | Silc Technologies, Inc. | Separation of light signals in a LIDAR system |
| US20230288566A1 (en) * | 2022-03-09 | 2023-09-14 | Silc Technologies, Inc. | Adjusting imaging system data in response to edge effects |
| US12422618B2 (en) | 2022-10-13 | 2025-09-23 | Silc Technologies, Inc. | Buried taper with reflecting surface |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1131789A1 (fr) * | 1999-09-17 | 2001-09-12 | THOMSON multimedia | Procede de construction d'un modele de scene 3d par analyse de sequence d'images |
| WO2002069265A1 (fr) * | 2001-02-23 | 2002-09-06 | Holding B.E.V. S.A. | Procede et dispositif de reperage d'un objet par sa forme, sa dimension et/ou son orientation |
| FR2858447A1 (fr) | 2003-07-29 | 2005-02-04 | Holding Bev Sa | Procede et dispositif automatise de perception avec determination et caracterisation de bords et de frontieres d'objets d'un espace, construction de contours et applications |
| US20050097068A1 (en) | 2003-10-29 | 2005-05-05 | Thore Graepel | Invariant pattern recognition |
| US20120308136A1 (en) | 2010-03-26 | 2012-12-06 | Izhikevich Eugene M | Apparatus and methods for pulse-code invariant object recognition |
| US20150227849A1 (en) | 2009-12-07 | 2015-08-13 | Yahoo! Inc. | Method and System for Invariant Pattern Recognition |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2540263B1 (fr) * | 1983-01-31 | 1988-05-20 | Commissariat Energie Atomique | Procede de reconnaissance automatique d'une image a partir d'une image correspondante de reference |
| FR2751772B1 (fr) * | 1996-07-26 | 1998-10-16 | Bev Bureau Etude Vision Soc | Procede et dispositif fonctionnant en temps reel, pour le reperage et la localisation d'une zone en mouvement relatif dans une scene, ainsi que pour la determination de la vitesse et la direction du deplacement |
| FR2884625A1 (fr) | 2005-04-15 | 2006-10-20 | Holding Bev Sa Sa Luxembourgeo | Procede de guidage automatique de vehicule dans une voie de circulation, dispositif correspondant |
| CN102999547A (zh) * | 2011-09-16 | 2013-03-27 | 莱克西私人有限公司 | 搜索查询的自动生成 |
| CN103268498B (zh) * | 2013-04-21 | 2016-06-29 | 郑州轻工业学院 | 一种感兴趣区域模糊图像语义理解的方法 |
| FR3039684B1 (fr) | 2015-07-27 | 2018-08-10 | Compagnie Generale Des Etablissements Michelin | Procede optimise d'analyse de la conformite de la surface d'un pneumatique |
| CN106126590A (zh) * | 2016-06-20 | 2016-11-16 | 北京航空航天大学 | 一种基于关键信息的无人机侦察视频检索方法 |
-
2017
- 2017-04-28 FR FR1700465A patent/FR3065825B1/fr active Active
-
2018
- 2018-04-27 KR KR1020197028748A patent/KR102454841B1/ko active Active
- 2018-04-27 JP JP2019554641A patent/JP6909307B2/ja active Active
- 2018-04-27 CN CN201880025028.7A patent/CN110520869B/zh active Active
- 2018-04-27 WO PCT/EP2018/060929 patent/WO2018197693A1/fr not_active Ceased
- 2018-04-27 US US16/499,820 patent/US11164049B2/en active Active
- 2018-04-27 EP EP18718863.6A patent/EP3616132A1/fr not_active Ceased
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1131789A1 (fr) * | 1999-09-17 | 2001-09-12 | THOMSON multimedia | Procede de construction d'un modele de scene 3d par analyse de sequence d'images |
| WO2002069265A1 (fr) * | 2001-02-23 | 2002-09-06 | Holding B.E.V. S.A. | Procede et dispositif de reperage d'un objet par sa forme, sa dimension et/ou son orientation |
| FR2858447A1 (fr) | 2003-07-29 | 2005-02-04 | Holding Bev Sa | Procede et dispositif automatise de perception avec determination et caracterisation de bords et de frontieres d'objets d'un espace, construction de contours et applications |
| US20050097068A1 (en) | 2003-10-29 | 2005-05-05 | Thore Graepel | Invariant pattern recognition |
| US20150227849A1 (en) | 2009-12-07 | 2015-08-13 | Yahoo! Inc. | Method and System for Invariant Pattern Recognition |
| US20120308136A1 (en) | 2010-03-26 | 2012-12-06 | Izhikevich Eugene M | Apparatus and methods for pulse-code invariant object recognition |
Non-Patent Citations (3)
| Title |
|---|
| "Perceptive Invariance and Associative Memory Between Perception and Semantic Representation, USER a Universal Semantic Représentation Implemented in a System on Chip (SoC", LIVING MACHINES 2016, LNAI, vol. 9793, 2016, pages 275 - 287 |
| PIRIM PATRICK ED - WALTER DIDIMO ET AL: "Perceptive Invariance and Associative Memory Between Perception and Semantic Representation USER a Universal SEmantic Representation Implemented in a System on Chip (SoC)", 12 July 2016, NETWORK AND PARALLEL COMPUTING; [LECTURE NOTES IN COMPUTER SCIENCE; LECT.NOTES COMPUTER], SPRINGER INTERNATIONAL PUBLISHING, CHAM, PAGE(S) 275 - 287, ISBN: 978-3-540-28012-5, ISSN: 0302-9743, XP047350737 * |
| See also references of EP3616132A1 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110520869A (zh) | 2019-11-29 |
| FR3065825A1 (fr) | 2018-11-02 |
| US20200104647A1 (en) | 2020-04-02 |
| US11164049B2 (en) | 2021-11-02 |
| JP6909307B2 (ja) | 2021-07-28 |
| KR102454841B1 (ko) | 2022-10-13 |
| JP2020519984A (ja) | 2020-07-02 |
| CN110520869B (zh) | 2023-06-02 |
| FR3065825B1 (fr) | 2022-12-23 |
| EP3616132A1 (fr) | 2020-03-04 |
| KR20190122790A (ko) | 2019-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3616132A1 (fr) | Procédé et dispositif automatisés aptes à assurer l'invariance perceptive d'un évènement spatio-temporel dynamiquement en vue d'en extraire des représentations sémantiques unifiées | |
| US20190073553A1 (en) | Region proposal for image regions that include objects of interest using feature maps from multiple layers of a convolutional neural network model | |
| CN110516201A (zh) | 图像处理方法、装置、电子设备及存储介质 | |
| US12217474B2 (en) | Detection of moment of perception | |
| CN110210513A (zh) | 数据分类方法、装置及终端设备 | |
| EP1259939B1 (fr) | Procede et dispositif de perception automatique | |
| US20230039592A1 (en) | Image sensor with integrated efficient multiresolution hierarchical deep neural network (dnn) | |
| KR102510881B1 (ko) | 신경망 기반의 특징점 학습 장치 및 방법 | |
| EP2364490A1 (fr) | Dispositif a architecture pipeline de flot de donnees pour la reconnaissance et la localisation d'objets dans une image par balayage de fenetres de detection | |
| US11915419B1 (en) | Auto-normalization for machine learning | |
| Shende et al. | Using deep learning to detect deepfake videos | |
| EP2839410A1 (fr) | Procede de reconnaissance d'un contexte visuel d'une image et dispositif correspondant | |
| EP3198523A1 (fr) | Procede et systeme de detection d'evenements de nature connue | |
| Areeb et al. | AI art critic: Artistic classification of poster images using neural networks | |
| CN115705758A (zh) | 活体识别方法、装置、电子设备和存储介质 | |
| CN117391201A (zh) | 问答方法、装置及电子设备 | |
| WO2024102521A1 (fr) | Attention parmi des vues pour tâches de perception visuelle à l'aide de multiples entrées de caméra | |
| Liu et al. | Fine-art recognition using convolutional transformers | |
| FR2821459A1 (fr) | Procede et dispositif de reperage d'un objet par sa forme, sa dimension et/ou son orientation | |
| WO2018197687A1 (fr) | Dispositif ram automatisé apte à mémoriser, rappeler et gérer de manière non volatile, des associations de messages versus label et vice versa, avec un maximum de vraisemblance | |
| US20250095259A1 (en) | Avatar animation with general pretrained facial movement encoding | |
| Kabilan et al. | Advanced Temporal Analysis for Deepfake Detection using XceptionNet in Media Forensics | |
| WO2025106448A1 (fr) | Synthèse de vues à l'aide d'un modèle d'apprentissage automatique de génération d'image multi-vues | |
| Dongre et al. | A STUDY ON DEEP LEARNING MODELS AND ALGORITHMS USED IN COMPUTER VISION | |
| WO2025106450A1 (fr) | Modification de contenu vidéo |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18718863 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 20197028748 Country of ref document: KR Kind code of ref document: A Ref document number: 2019554641 Country of ref document: JP Kind code of ref document: A |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2018718863 Country of ref document: EP |
|
| ENP | Entry into the national phase |
Ref document number: 2018718863 Country of ref document: EP Effective date: 20191128 |