Abstract: The present disclosure provides a method for detecting a defect of a film. The method includes obtaining a film image, determining one or more pieces of scratch information corresponding to the film image through processing the film image using a recognition model, the recognition model includes a convolution layer, a regression layer, and a classification layer, determining whether each piece of scratch information in the one or more pieces of scratch information meets a preset condition, each piece of scratch information includes position information, angle information, and size information, in response to a determination that each piece of scratch information meets the preset condition, adding one or more pieces of annotation information to the one or more pieces of scratch information that meets the preset condition, and generating prompt information based on the one or more pieces of annotation information.

Abstract: The present disclosure provides a method for recognizing a small target based on a deep learning network. The method comprises: determining, based on a collected image, spot defect information through a recognition model including a first feature determination layer, a second feature determination layer, and a spot defect determination layer, determining, based on the collected image, a first feature map, determining, based on the first feature map, a second feature map by fusing with positional encoding, determining, based on the second feature map, a third feature map through the second feature determination layer, and obtaining, based on the third feature map, positional information of the spot defect through a first determination layer, and determining, based on the third feature map, classification information of the spot defect through a second determination layer.

Abstract: The present disclosure provides a method for recognizing a small target based on a deep learning network. The method comprises: determining, based on a collected image, spot defect information through a recognition model including a first feature determination layer, a second feature determination layer, and a spot defect determination layer, determining, based on the collected image, a first feature map, determining, based on the first feature map, a second feature map by fusing with positional encoding, determining, based on the second feature map, a third feature map through the second feature determination layer, and obtaining, based on the third feature map, positional information of the spot defect through a first determination layer, and determining, based on the third feature map, classification information of the spot defect through a second determination layer.

Abstract: An end-to-end multimodal gait recognition method based on deep learning includes: first extracting gait appearance features (color, texture and the like) through RGB video frames, and obtaining a mask by semantic segmentation of the RGB video frames; then extracting gait mask features (contour and the like) through the mask; and finally performing fusion and recognition on the two kinds of features. The method is configured for extracting gait appearance feature and mask feature by improving GaitSet, improving semantic segmentation speed on the premise of ensuring accuracy through simplified FCN, and fusing the gait appearance feature and the mask feature to obtain a more complete information representation.

Abstract: The present disclosure provides a method for detecting a defect of a film. The method includes obtaining a film image, determining one or more pieces of scratch information corresponding to the film image through processing the film image using a recognition model, the recognition model includes a convolution layer, a regression layer, and a classification layer, determining whether each piece of scratch information in the one or more pieces of scratch information meets a preset condition, each piece of scratch information includes position information, angle information, and size information, in response to a determination that each piece of scratch information meets the preset condition, adding one or more pieces of annotation information to the one or more pieces of scratch information that meets the preset condition, and generating prompt information based on the one or more pieces of annotation information.

Abstract: An end-to-end multimodal gait recognition method based on deep learning includes: first extracting gait appearance features (color, texture and the like) through RGB video frames, and obtaining a mask by semantic segmentation of the RGB video frames; then extracting gait mask features (contour and the like) through the mask; and finally performing fusion and recognition on the two kinds of features. The method is configured for extracting gait appearance feature and mask feature by improving GaitSet, improving semantic segmentation speed on the premise of ensuring accuracy through simplified FCN, and fusing the gait appearance feature and the mask feature to obtain a more complete information representation.

Abstract: A stereoscopic video generation method based on 3D convolution neural network is disclosed, which is able to convert existing 2D video sources into stereoscopic videos. The method includes preparing the training data, dividing the training video sources into left eye view sequences and right eye view sequences; and then processing the left eye image sequences through shot segmentation via fuzzy C-means clustering method, calculating a mean image of all left eye images, subtracting the mean image from the left eye images, taking the right eye images as a training target; training the obtained 3D convolution neural network through the training data; processing the 2D video sources which need to be converted into stereoscopic videos in the same way as training set, inputting to the trained 3D convolution neural network to obtain the right eye view image sequences of the 2D videos; and finally combining the two to be stereoscopic videos.

Abstract: A stereoscopic video generation method based on 3D convolution neural network is disclosed, which is able to convert existing 2D video sources into stereoscopic videos. The method includes preparing the training data, dividing the training video sources into left eye view sequences and right eye view sequences; and then processing the left eye image sequences through shot segmentation via fuzzy C-means clustering method, calculating a mean image of all left eye images, subtracting the mean image from the left eye images, taking the right eye images as a training target; training the obtained 3D convolution neural network through the training data; processing the 2D video sources which need to be converted into stereoscopic videos in the same way as training set, inputting to the trained 3D convolution neural network to obtain the right eye view image sequences of the 2D videos; and finally combining the two to be stereoscopic videos.

Abstract: A depth estimation method for a monocular image based on a multi-scale CNN and a continuous CRF is disclosed in this invention. A CRF module is adopted to calculate a unary potential energy according to the output depth map of a DCNN, and the pairwise sparse potential energy according to input RGB images. MAP (maximum a posteriori estimation) algorithm is used to infer the optimized depth map at last. The present invention integrates optimization theories of the multi-scale CNN with that of the continuous CRF. High accuracy and a clear contour are both achieved in the estimated depth map; the depth estimated by the present invention has a high resolution and detailed contour information can be kept for all objects in the scene, which provides better visual effects.

Abstract: A depth estimation method for a monocular image based on a multi-scale CNN and a continuous CRF is disclosed in this invention. A CRF module is adopted to calculate a unary potential energy according to the output depth map of a DCNN, and the pairwise sparse potential energy according to input RGB images. MAP (maximum a posteriori estimation) algorithm is used to infer the optimized depth map at last. The present invention integrates optimization theories of the multi-scale CNN with that of the continuous CRF. High accuracy and a clear contour are both achieved in the estimated depth map; the depth estimated by the present invention has a high resolution and detailed contour information can be kept for all objects in the scene, which provides better visual effects.