Huazhong University of Science and Technology

You Yang

Qiong Liu

Jun Zhang

Qian Kan

Ruofei Xiong

Xinshengzi Huang

The present invention discloses a scene recognition method, a training method and a device based on pyramid attention, belonging to the field of computer vision. The method includes: pyramid layering a color feature map and a depth feature map respectively, calculating the corresponding attention map of each layer; taking the output of the attention of the last layer as the output; taking the attention output of the last layer as the final feature map, for the remaining layers, adding the result after upsampling of the final feature map of an upper layer to the attention output of this layer as the final feature map of this layer; scaling the attention map and the final feature map, using the average of two new attention maps as the final attention map, mapping the largest k position in the final attention map to the final feature map of this layer.

Scene recognition method, training method and device based on pyramid attention

The present invention discloses a pedestrian trajectory prediction method and system based on a multi-interaction spatio-temporal graph network, which belong to the field of pedestrian trajectory prediction. The method includes: extracting multi-interaction features of pedestrians in video frames; for each frame in a video sequence, abstracting each pedestrian in the frame as a vertex, connecting the pedestrian(s) with other pedestrians to form an edge(s), where the vertex attribute is the multi-interaction feature of the pedestrian so as to obtain a multi-interaction spatio-temporal graph network; for each multi-interaction spatio-temporal graph, obtaining spatial dependencies of each pedestrian with other pedestrians in the spatio-temporal graph, and optimizing the attribute of each vertex through the spatial dependencies between pedestrians; the vertices of adjacent time points of the pedestrians are connected to obtain time dependencies of the pedestrians, and then infer a trajectory of a future time point(s).

Pedestrian trajectory prediction method and system based on multi-interaction spatiotemporal graph network

A point cloud reconstruction method and apparatus based on a pyramid transformer, a device, and a medium are provided. The method includes: after obtaining point feature data corresponding to each view, inputting the point feature data into a dual-channel pyramid network; when passing through a first channel, inputting an updated output of an attention mechanism of each layer to a next layer, when passing through a second channel, splicing and inputting the updated output of the attention mechanism of each layer and the updated output of the attention mechanism of a corresponding layer of the first channel in a same layer as the next layer to the next layer; converting the updated output of the attention mechanism of a bottommost layer of the second channel into point cloud coordinates to obtain a point cloud result, and taking an intersection of the point cloud result to characterize a target object.

Growing community of inventors

Hubei, China

You Yang