torch.load("/home/turgutluk/data/models/chx-mimic-only-pneumo-densenet121-320.pth")

KeremTurgutlu / siim-acr-pneumothorax

Kaggle competition repo: https://www.kaggle.com/c/siim-acr-pneumothorax-segmentation/

MIT License

4 stars 1 forks source link

torch.load("/home/turgutluk/data/models/chx-mimic-only-pneumo-densenet121-320.pth") #1

Open sugerpopo opened 4 years ago

sugerpopo commented 4 years ago

Hi, @KeremTurgutlu I'm trying to run densenet121+unet on fastai to do semantic segmentation and find your code can support it. But ,I don't know where your densenet model come from.

chx_state_dict = torch.load("/home/turgutluk/data/models/chx-mimic-only-pneumo-densenet121-320.pth")

Could you help me Please

KeremTurgutlu commented 4 years ago

I don't push weights and large files in general to github. Also, in my experience densenet + unet doesn't give good results probably because of the dense connections in the architecture or me not being able to find a good split. Instead, try efficientnets + unet.