search

WeijingShi / Point-GNN

Point-GNN: Graph Neural Network for 3D Object Detection in a Point Cloud, CVPR 2020.
MIT License
523 stars 114 forks source link

Evaluation Results #6

Closed dirk61 closed 4 years ago

dirk61 commented 4 years ago

I ran eval.py on both car and ped-cyclist, the results are as followed: image image I wonder how can I get mAP for each difficulty level(Easy, Moderate, Hard) respectively, which is shown in Table 2. in your paper.

WeijingShi commented 4 years ago

The eval.py is meant to be a monitoring task for the training process. It does not output the KITTI APs. For that, you need to use the run.py instead and use the native_kitti_evaluation tool. The readme file has the instruction on how to use them. For the test set results in Table 2, you need to submit the results to the KITTI server, as we do not have the ground truth.