search

DerrickXuNu / OpenCOOD

[ICRA 2022] An opensource framework for cooperative detection. Official implementation for OPV2V.
https://mobility-lab.seas.ucla.edu/opv2v/
Other
644 stars 99 forks source link

Something about the shrink_features? #103

Closed lubin2022 closed 1 year ago

lubin2022 commented 1 year ago

Thanks for you excellent work before, I have a question about the shrink_features. As we know, before the fision, the features will go through the shrink_conv module to reduce memory. The shape of the tensor after the shrink_conv module are (N, C, H, W), the question is that what's the N satand for? As I set the max_cav to 5 in the yaml file, but I found that the N will be larger than 5. So I will appreciate it if you can explain that.

DerrickXuNu commented 1 year ago

Thanks for using Opencood. The N is actually batch_size multiplied with number of agents, which can be larger than 5. This dimension will be split into (B, L) later.