SEATTLE - TRI’s Machine Learning team will present four different papers this week at the 2020 Conference on Computer Vision and Pattern Recognition (CVPR). The annual premier computer vision event will be held virtually from June 13 to June 19.  Of the four papers, three will be oral presentations, which is a distinct honor at this highly selective event. This year, CVPR received 6,656 paper submissions and accepted only 1470 (22%). Out of these accepted papers, only the 335 highest rated papers (5%) were selected for oral presentations.  The oral presentations are for:

• "3D Packing for Self-Supervised Monocular Depth Estimation"

• "Autolabeling 3D Objects with Differentiable Rendering of SDF Shape Priors"

• "Real-Time Panoptic Segmentation from Dense Detections"

The fourth paper accepted at CVPR is "Spatio-Temporal Graph for Video Captioning with Knowledge Distillation."

Event Link