Data Heterogeneity - Githubissues

AIR-THU / DAIR-RCooper

[CVPR2024] Official implementation of "RCooper: A Real-world Large-scale Dataset for Roadside Cooperative Perception"

https://openaccess.thecvf.com/content/CVPR2024/html/Hao_RCooper_A_Real-world_Large-scale_Dataset_for_Roadside_Cooperative_Perception_CVPR_2024_paper.html

55 stars 5 forks source link

Data Heterogeneity #8

Closed githigher closed 2 months ago

githigher commented 2 months ago

In intersection scenes, you configure two types of LiDARs, Multiline LiDARs Group and MEMS LiDARs. I would like to know if it is possible to get the type of LIDAR features in the code.😊

leofansq commented 2 months ago

Towards the unified cooperative perception for practical application, we didn't distinguish the types in the code. However, you can get the type according to the infrastructure's IDs.

Multiline LiDAR: 117, 120, 136, 139 MEMS LiDAR: 118, 119, 137, 138

githigher commented 1 month ago

Thank you very much for your reply! Can you be more specific as to where the code to get the infrastructure ID is located? I'd like to try two different encoders for Multiline LiDARs Group and MEMS LiDARs.

leofansq commented 1 month ago

For the raw RCooper dataset, you can directly distinguish the ID in data folder.
For the converted OPV2V-format or V2V4Real-format dataset, you can distinguish the infrastructures according to the following mapping relationship
- '116-115' scenario: 116 ---> 0, 115 ---> 1
- '106-105' scenario: 106 ---> 0, 105 ---> 1
- '117-118-120-119' scenario: 117 ---> 0, 118 ---> 1, 120 ---> 2, 119 ---> 3
- '136-137-138-139' scenario: 136 ---> 0, 137 ---> 1, 138 ---> 2, 139 ---> 3