Skip to content

Latest commit

 

History

History
111 lines (67 loc) · 2.84 KB

perception_eval.md

File metadata and controls

111 lines (67 loc) · 2.84 KB

Multi-Ego Cooperative Perception

OPV2V 3D Object Detection

CoBEVT, 100ms Delay, no compression

python opencood/tools/multi_ego_inference_opv2v.py --perception_model_name corpbevtlidar_delay_1_frame_aug

CoBEVT, 100ms Delay, 256x compression

python opencood/tools/multi_ego_inference_opv2v.py --perception_model_name corpbevtlidar_delay_1_frame_aug_c256

V2VNet, 100ms Delay, no compression

python opencood/tools/multi_ego_inference_opv2v.py --perception_model_name point_pillar_v2vnet_multiego

No Cooperation

python opencood/tools/multi_ego_inference_opv2v.py --perception_model_name point_pillar_sinbevt --no_coop

This will save the detection and tracking results in the preprocessed_data folder. This script caches intermediate results in the preprocessed_data folder as well, so if you want to re-run, you need to delete the cache. This script also saves the fused features in the preprocessed_data folder. Our intermediate results can be accessed HERE.

The overall evaluation results will be saved in the checkpoint folder eval.yaml.

V2V4Real 3D Object Detection

CoBEVT, 100ms Delay, no compression

python opencood/tools/multi_ego_inference_v2v4real.py --perception_model_name point_pillar_cobevt_multiego_1x

CoBEVT, 100ms Delay, 256x compression

python opencood/tools/multi_ego_inference_v2v4real.py --perception_model_name point_pillar_cobevt_multiego_256x

V2VNet, 100ms Delay, no compression

python opencood/tools/multi_ego_inference_v2v4real.py --perception_model_name point_pillar_v2vnet_multiego

No Cooperation

python opencood/tools/multi_ego_inference_v2v4real.py --perception_model_name point_pillar_sinbevt --no_coop

OPV2V 3D Multi-Object Tracking

CoBEVT, 100ms Delay, no compression

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name corpbevtlidar_delay_1_frame_aug --dataset opv2v

CoBEVT, 100ms Delay, 256x compression

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name corpbevtlidar_delay_1_frame_aug_c256 --dataset opv2v

No Cooperation

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name point_pillar_sinbevt --dataset opv2v

V2V4Real 3D Multi-Object Tracking

CoBEVT, 100ms Delay, no compression

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name point_pillar_cobevt_multiego_1x --dataset v2v4real

CoBEVT, 100ms Delay, 256x compression

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name point_pillar_cobevt_multiego_256x --dataset v2v4real

No Cooperation

python AB3Dmot/scripts/KITTI/evaluate.py --perception_model_name point_pillar_sinbevt --dataset v2v4real