Submission details of MMPAT_CVPR21

Paper Link
Code Link N/A
AP 0.678829
Input N/A
Runtime 0.07 s
Environment 1 GPU (Titan X)
Abstract Panoramic image detection and tracking using both image and point clouds. For detection, we adopt the Cascade rcnn as our object detector. The backbone of the detector is ResNet 101. The detector is pre-trained on the COCO dataset and then fine tuned on the JRDB dataset. During training, we employ data augment methods such as mixup, random crop, multiscale training to augment the dataset. During inference, we employ horizontal flip and softnms techniques to generate more robust resutls.

Detailed results

Overall precision/recall curve

Per-sequence results

sequence name AP
cubberly-auditorium-2019-04-22_1 0.733309
discovery-walk-2019-02-28_0 0.825952
discovery-walk-2019-02-28_1 0.803498
food-trucks-2019-02-12_0 0.789975
gates-ai-lab-2019-04-17_0 0.758997
gates-basement-elevators-2019-01-17_0 0.798955
gates-foyer-2019-01-17_0 0.835772
gates-to-clark-2019-02-28_0 0.760957
hewlett-class-2019-01-23_0 0.861023
hewlett-class-2019-01-23_1 0.907767
huang-2-2019-01-25_1 0.632578
huang-intersection-2019-01-22_0 0.742469
indoor-coupa-cafe-2019-02-06_0 0.633404
lomita-serra-intersection-2019-01-30_0 0.719845
meyer-green-2019-03-16_1 0.502327
nvidia-aud-2019-01-25_0 0.6284
nvidia-aud-2019-04-18_1 0.784675
nvidia-aud-2019-04-18_2 0.781549
outdoor-coupa-cafe-2019-02-06_0 0.53751
quarry-road-2019-02-28_0 0.713112
serra-street-2019-01-30_0 0.673061
stlc-111-2019-04-19_1 0.834388
stlc-111-2019-04-19_2 0.835003
tressider-2019-03-16_2 0.586687
tressider-2019-04-26_0 0.611034
tressider-2019-04-26_1 0.674612
tressider-2019-04-26_3 0.644879
total 0.678829