search

wayveai / fiery

PyTorch code for the paper "FIERY: Future Instance Segmentation in Bird's-Eye view from Surround Monocular Cameras"
https://wayve.ai/blog/fiery-future-instance-prediction-birds-eye-view
MIT License
556 stars 85 forks source link

There are a lot of detected instance centers #30

Open Linkcy97 opened 2 years ago

Linkcy97 commented 2 years ago

When I training nuscense dataset，it display issue ：There are a lot of detected instance centers: torch.Size([4377, 2])

kaanakan commented 2 years ago

Hi,

Does it happen at the beginning of the training (in the validation sanity check)? If so, I think it is expected because the model is not trained, and it makes some random predictions. However, if it happens after some training, it probably means that your training has failed because after a few epochs model converges into a point where it makes reasonable predictions.

P.S.: I'm not an author of the paper; I just know the code and the repository.

Linkcy97 commented 2 years ago

Thank you for reply！