Techniques for spatially preserving flattening in deep learning neural networks are provided. In one aspect, a spatially preserving flattening module includes: a predictor for generating image feature maps from at least one convolutional layer of a feature extraction phase of a deep learning neural network applied to input image data; an auto-encoder for producing encodings of the image feature maps that preserve location and shape information associated with objects in the input image data; and a flattener for concatenating the encodings of the image feature maps to form a spatially preserving flattened encoding vector. A deep learning neural network that includes the present spatially preserving flattening module is also provided, as is a method for spatially preserving flattening.

Disclosed is an image analysis method including inputting analysis data, including information regarding an analysis target cell to a deep learning algorithm having a neural network structure, and analyzing an image by calculating, by use of the deep learning algorithm, a probability that the analysis target cell belongs to each of morphology classifications of a plurality of cells belonging to a predetermined cell group.

The invention provides a method for intelligently measuring vehicle trajectory based on a binocular stereo vision system, including the following steps: inputting a dataset into an SSD (Single Shot Multibox Detector) neural network to train a license plate recognition model; calibrating the binocular stereo vision system, and recording videos of moving target vehicles; detecting the license plates in the video frames with the license plate recognition model; performing stereo matching on the license plates in the subsequent frames of the same camera and in the corresponding left-view and right-view video frames by a feature-based matching algorithm; reserving correct matching point pairs after filtering with a homography matrix; screening the reserved matching point pairs, and reserving the one closest to the license plate center as the position of the target vehicle in the current frame; performing stereo measurement on the screened and reserved matching point pairs to get the spatial position coordinates of the vehicle in the video frames; and generating the moving trajectory of the vehicle in time sequence. The present invention is easy to install and adjust, and can simultaneously measure multiple target vehicles in multiple directions and on multiple lanes.

Systems, apparatuses, and methods for converting data to a tiling format when implementing convolutional neural networks are disclosed. A system includes at least a memory, a cache, a processor, and a plurality of compute units. The memory stores a first buffer and a second buffer in a linear format, where the first buffer stores convolutional filter data and the second buffer stores image data. The processor converts the first and second buffers from the linear format to third and fourth buffers, respectively, in a tiling format. The plurality of compute units load the tiling-formatted data from the third and fourth buffers in memory to the cache and then perform a convolutional filter operation on the tiling-formatted data. The system generates a classification of a first dataset based on a result of the convolutional filter operation.

The disclosure extends to methods, systems, and apparatuses for automated fixation generation and more particularly relates to generation of synthetic saliency maps. A method for generating saliency information includes receiving a first image and an indication of one or more sub-regions within the first image corresponding to one or more objects of interest. The method includes generating and storing a label image by creating an intermediate image having one or more random points. The random points have a first color in regions corresponding to the sub-regions and a remainder of the intermediate image having a second color. Generating and storing the label image further includes applying a Gaussian blur to the intermediate image.

Disclosed is an image analysis method including inputting analysis data, including information regarding an analysis target cell to a deep learning algorithm having a neural network structure, and analyzing an image by calculating, by use of the deep learning algorithm, a probability that the analysis target cell belongs to each of morphology classifications of a plurality of cells belonging to a predetermined cell group.

An image processing method, an image processing device, a training method and a computer-readable storage medium. The image processing method includes: extracting a characteristic vector in an image to be recognized; based on the characteristic vector of the image to be recognized, acquiring a predicted score value of the image to be recognized; and based on the predicted score value, determining a category of an image information of the image to be recognized; wherein the image to be recognized is a face image, and the image information is a facial expression.

In a state where a tracking flag is on, a search controller determines, with respect to a previous frame, whether an amount of change in positional coordinates of a feature point of a face in the current frame is within a predetermined range, whether an amount of change in face orientation is within a predetermined angle range, and whether an amount of change in sight line direction is within a predetermined range. When the conditions are satisfied in all these determinations, the change in the detection result in the current frame with respect to the previous frame is considered as being within an allowable range, and continuously in a subsequent frame, detection processing for a face image is performed in accordance with a face image area saved in a tracking information storage unit.

A method includes acquiring, from a camera, an image data sequence of a real object in a real scene and performing a first template-matching on an image frame in the image data sequence using intensity-related data sets stored in one or more memories to generate response maps. The intensity-related data sets represent an intensity distribution of a reference object from respective viewpoints. The reference object corresponds to the real object. A candidate region of interest is determined for the real object in the image frame based on the response maps, and second template-matching is performed on the candidate region of interest using shape-related feature data sets stored in one or more memories to derive a pose of the real object. The shape-related feature data sets represent edge information of the reference object from the respective viewpoints.

A computer implemented method for automatic print analysis, the method comprising: receiving a first image wherein the first image shows one or more of: a latent print, a patent print, an impressed print, and an actual finger, palm, toe and/or foot; and wherein the first image includes characteristic features of at least one of a finger, a palm, a toe and a foot; creating an orientation field by estimating the orientation of one or more features in the first image, wherein the estimating comprises: applying an orientation operator to the first image, the orientation operator being based on a plurality of isotropic filters lying in quadrature.