ucla-mobility / OpenCDA

A generalized framework for prototyping full-stack cooperative driving automation applications under CARLA+SUMO.
Other
988 stars 197 forks source link

Coperception - support cooperative perception based on OpenCOOD #187

Open ju40268 opened 1 year ago

ju40268 commented 1 year ago

Scenario v2xp_online_carla.py Run with command

python opencda.py -t v2xp_online_carla -v 0.9.12 --apply_ml 

No Comm

Early:
The Average Precision at IOU 0.3 is 0.41, The Average Precision at IOU 0.5 is 0.37, The Average Precision at IOU 0.7 is 0.29
Intermediate
The Average Precision at IOU 0.3 is 0.48, The Average Precision at IOU 0.5 is 0.40, The Average Precision at IOU 0.7 is 0.28
Late:
The Average Precision at IOU 0.3 is 0.44, The Average Precision at IOU 0.5 is 0.38, The Average Precision at IOU 0.7 is 0.31
DerrickXuNu commented 1 year ago

A quick question: did you change any code in opencood?

ju40268 commented 1 year ago

The parameters relate to cooperative perception is listed with parameters within the yaml file as follows:

coperception:
  models:
    early: 'opencood/logs/pointpillar_early_fusion'
    intermediate_v2xvit: 'opencood/logs/pointpillar_v2xvit_fusion'
    intermediate_attentive: 'opencood/logs/pointpillar_attentive_fusion'
    intermediate_cobevt: 'opencood/logs/pointpillar_cobevt_fusion'
    late: 'opencood/logs/pointpillar_late_fusion'
  fusion_method: 'intermediate_v2xvit'

and whether or not to enable communication across cav, and plotting ground truth is by the config

coperception: true # cooperative perception
enable_communicate: true
enable_show_gt: false

under the perception section. Example evaluation results for intermediate fusion using v2xvit

Evaluate final average precision results:
  - Fusion method: intermediate_v2xvit
  - The Average Precision at IOU 0.3 is 0.66, The Average Precision at IOU 0.5 is 0.57, The Average Precision at IOU 0.7 is 0.36
DerrickXuNu commented 1 year ago

Is this ready to merge?