title = {Semantic Amodal Segmentation}, booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {July}, year = {2017} } Deep Sequential Context Networks for Action Prediction Yu Kong, Zhiqiang Tao, Yun Fu [pdf] [bibtex] @InProc...
Compared to pointpillars10, this method has received great increase in mAP, some categories such as traffic corns, pedestrian and bicycle often have few LiDAR points on them, thus the additional appearance features provided by color and texture information is extremely valuable. The reason that ...
Among the prior amodal object completion works with appearance recovery, Ehsani et al [8] generate the occluded parts of objects us- ing Unet [49] by leveraging about 5,000 synthetic images restricted to indoor scenes such as kitchen and living room. Yan et al [9] recover the appearance ...
Instead of following appearance attributes using 2D bounding boxes (BBs), it computes the position of targets in the 3D world using ge- ometry contained in 3D BBs. 3D object tracking either focuses on RGB-D information [48], by mimicking the 2D object tracking methods but with a...