search

bradyz / cross_view_transformers

Cross-view Transformers for real-time Map-view Semantic Segmentation (CVPR 2022 Oral)
MIT License
531 stars 81 forks source link

Model parameter count #14

Closed vztu closed 2 years ago

vztu commented 2 years ago

Hi Brady, thanks for the great work!

I have a question regarding the model parameters. In your paper, you have reported 5M in Table 1 on nuscenes. However, I run your code with the default yaml config, and the log says 1.1M parameters. Did I miss anything?

image

bradyz commented 2 years ago

At the time of the paper we had included weights of the entire backbone (efficientnet), when much of the later layers weren't actually used. So after culling those layers, this 1.1M params is the correct number.

We'll fix the paper's numbers in a future revision, thanks for bringing this up 👍