Submission details of EPNet

Name EPNet
Paper Link
Code Link
AP 0.635886
Input N/A
Runtime 0.23 s
Environment 1 GPU(Titan V)
Abstract In this paper, we aim at addressing two critical issues in the 3D detection task, including the exploitation of multiple sensors (namely LiDAR point cloud and camera image), as well as the inconsistency between the localization and classification confidence.To this end, we propose a novel fusion module to enhance the point features with semantic image features in a point-wise manner without any image annotations. Besides, a consistency enforcing loss is employed to explicitly encourage the consistency of both the localization and classification confidence.

Detailed results

Overall precision/recall curve

Per-sequence results

sequence name AP
cubberly-auditorium-2019-04-22_1 0.762286
discovery-walk-2019-02-28_0 0.90068
discovery-walk-2019-02-28_1 0.821669
food-trucks-2019-02-12_0 0.825449
gates-ai-lab-2019-04-17_0 0.758043
gates-basement-elevators-2019-01-17_0 0.86038
gates-foyer-2019-01-17_0 0.891658
gates-to-clark-2019-02-28_0 0.91943
hewlett-class-2019-01-23_0 0.783743
hewlett-class-2019-01-23_1 0.935491
huang-2-2019-01-25_1 0.669661
huang-intersection-2019-01-22_0 0.836635
indoor-coupa-cafe-2019-02-06_0 0.557868
lomita-serra-intersection-2019-01-30_0 0.944712
meyer-green-2019-03-16_1 0.633689
nvidia-aud-2019-01-25_0 0.649178
nvidia-aud-2019-04-18_1 0.653419
nvidia-aud-2019-04-18_2 0.808758
outdoor-coupa-cafe-2019-02-06_0 0.619114
quarry-road-2019-02-28_0 0.806433
serra-street-2019-01-30_0 0.869388
stlc-111-2019-04-19_1 0.904551
stlc-111-2019-04-19_2 0.872955
tressider-2019-03-16_2 0.815696
tressider-2019-04-26_0 0.512586
tressider-2019-04-26_1 0.492756
tressider-2019-04-26_3 0.550831
total 0.635886