Skip to content

Latest commit

 

History

History
23 lines (16 loc) · 2.08 KB

model_zoo.md

File metadata and controls

23 lines (16 loc) · 2.08 KB
Raw

Model Zoo

Detection

For the detection task we trained the CenterPoint network on the nuScenes dataset, with and without the velocity head, for 10 epochs each.

mAP mATE (m) mASE (1-IoU) mAOE (rad) mAVE (m/s) mAAE (1-acc) NDS checkpoint config
CenterPoint*1 0.480 0.308 0.264 0.409 1.193 0.446 0.497 model config.py
CenterPoint 0.469 0.311 0.268 0.432 0.388 0.197 0.575 model config.py

1: CenterPoint* indicates the CenterPoint variant trained only with the detection head

Further training is required to achieve the performance obtained by MMDetection3D

Tracking

For the tracking task we used the non-learning-based algorithm Kalman Filter, and a CenterPointTracker which computes trackings based on the output velocities of the CenterPoint detection network (with the velocity head). Thefore, we did not train a network for this task.

Prediction

For the prediction task we trained the LaneGCN network on the nuScenes dataset for 36 epochs.

MinADE_5 MinADE_10 MissRateTopK_2_5 MissRateTopK_2_10 MinFDE_1 OffroadRate checkpoint
LaneGCN 2.289 1,318 63.54% 50.74% 9.148 0.052 model