Submission details of T_HJ

Name T_HJ
Paper Link N/A
Code Link N/A
AP 0.680975
Input N/A
Runtime 0.1 s
Environment 1 GPU (Titan X)
Abstract Using the stitched image, we trained the type of R-CNN. 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 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 0.744256
discovery-walk-2019-02-28_0 0.829301
discovery-walk-2019-02-28_1 0.787673
food-trucks-2019-02-12_0 0.780853
gates-ai-lab-2019-04-17_0 0.743703
gates-basement-elevators-2019-01-17_0 0.875985
gates-foyer-2019-01-17_0 0.871777
gates-to-clark-2019-02-28_0 0.783403
hewlett-class-2019-01-23_0 0.896127
hewlett-class-2019-01-23_1 0.942881
huang-2-2019-01-25_1 0.644089
huang-intersection-2019-01-22_0 0.7388
indoor-coupa-cafe-2019-02-06_0 0.650763
lomita-serra-intersection-2019-01-30_0 0.668129
meyer-green-2019-03-16_1 0.511225
nvidia-aud-2019-01-25_0 0.64347
nvidia-aud-2019-04-18_1 0.791538
nvidia-aud-2019-04-18_2 0.809435
outdoor-coupa-cafe-2019-02-06_0 0.58879
quarry-road-2019-02-28_0 0.66704
serra-street-2019-01-30_0 0.616872
stlc-111-2019-04-19_1 0.82651
stlc-111-2019-04-19_2 0.820208
tressider-2019-03-16_2 0.61242
tressider-2019-04-26_0 0.605863
tressider-2019-04-26_1 0.672534
tressider-2019-04-26_3 0.646451
total 0.680975