Submission details of F-ConvNet

Name F-ConvNet
Paper Link
Code Link
AP 0.397807
Input N/A
Runtime 0.30 s
Environment 1 GPU (Gtx Titan X)
Abstract Abstract— In this work, we propose a novel method termed Frustum ConvNet (F-ConvNet) for amodal 3D object detection from point clouds. Given 2D region proposals in an RGB image, our method first generates a sequence of frustums for each region proposal, and uses the obtained frustums to group local points. F-ConvNet aggregates point-wise features as frustumlevel feature vectors, and arrays these feature vectors as a feature map for use of its subsequent component of fully convolutional network (FCN), which spatially fuses frustum-level features and supports an end-to-end and continuous estimation of oriented boxes in the 3D space. We also propose component variants of F-ConvNet, including an FCN variant that extracts multi-resolution frustum features, and a refined use of FConvNet over a reduced 3D space. Careful ablation studies verify the efficacy of these component variants. F-ConvNet assumes no prior knowledge of the working 3D environment and is thus dataset-agnostic. We present experiments on both the indoor SUN-RGBD and outdoor KITTI datasets. F-ConvNet outperforms all existing methods on SUN-RGBD, and at the time of submission it outperforms all published works on the KITTI benchmark.

Detailed results

Overall precision/recall curve

Per-sequence results

sequence name AP
cubberly-auditorium-2019-04-22_1 0.543881
discovery-walk-2019-02-28_0 0.503794
discovery-walk-2019-02-28_1 0.500886
food-trucks-2019-02-12_0 0.596174
gates-ai-lab-2019-04-17_0 0.476636
gates-basement-elevators-2019-01-17_0 0.465699
gates-foyer-2019-01-17_0 0.609531
gates-to-clark-2019-02-28_0 0.611199
hewlett-class-2019-01-23_0 0.525273
hewlett-class-2019-01-23_1 0.696575
huang-2-2019-01-25_1 0.433677
huang-intersection-2019-01-22_0 0.492944
indoor-coupa-cafe-2019-02-06_0 0.230959
lomita-serra-intersection-2019-01-30_0 0.694949
meyer-green-2019-03-16_1 0.381203
nvidia-aud-2019-01-25_0 0.315438
nvidia-aud-2019-04-18_1 0.435469
nvidia-aud-2019-04-18_2 0.517472
outdoor-coupa-cafe-2019-02-06_0 0.26504
quarry-road-2019-02-28_0 0.587019
serra-street-2019-01-30_0 0.429725
stlc-111-2019-04-19_1 0.70159
stlc-111-2019-04-19_2 0.722244
tressider-2019-03-16_2 0.555494
tressider-2019-04-26_0 0.278263
tressider-2019-04-26_1 0.384707
tressider-2019-04-26_3 0.363985
total 0.397807