search

Lkruitwagen / deepsentinel

DeepSentinel: a sentinel-1 and -2 self-supervised sensor fusion model for general purpose semantic embedding
43 stars 12 forks source link

How can I get the CLC_100m_colmap.json? #9

Open hukelin opened 3 years ago

hukelin commented 3 years ago

When I run your code,I have a question,How can I get the CLC_100m_colmap.json? Thank you!

Lkruitwagen commented 3 years ago

Hi! Apologies, I keep it in the data/ folder but I can see that that might not be helpful. Have added it to conf/: https://github.com/Lkruitwagen/deepsentinel/blob/master/conf/CLC_100m_colmap.json

hukelin commented 3 years ago

Thank you very much! By the way,when will you be able to provide a pre-training model?

Lkruitwagen commented 3 years ago

thanks for your interest! I'd like to release something a bit more 'official' whose performance I can vouch for, but I've just opened the experiments bucket for public read: https://console.cloud.google.com/storage/browser/deepsentinel-models. These are the model runs that give the results that can be found in the paper:

  VAE:
    resnet18: 9
    resnet18attn: 27
    resnet34: 20
  contrastive_loss:
    resnet18: 2
    resnet18attn: 28
    resnet34: 22
  none:
    resnet18: 3
    resnet18attn: 25
    resnet34: 21
  tile2vec:
    resnet18: 13
    resnet18attn: 26
    resnet34: 18

Let me know if you have access issues or any other questions!

NLei1993 commented 2 years ago

thanks for your interest! I'd like to release something a bit more 'official' whose performance I can vouch for, but I've just opened the experiments bucket for public read: https://console.cloud.google.com/storage/browser/deepsentinel-models. These are the model runs that give the results that can be found in the paper:

  VAE:
    resnet18: 9
    resnet18attn: 27
    resnet34: 20
  contrastive_loss:
    resnet18: 2
    resnet18attn: 28
    resnet34: 22
  none:
    resnet18: 3
    resnet18attn: 25
    resnet34: 21
  tile2vec:
    resnet18: 13
    resnet18attn: 26
    resnet34: 18

Let me know if you have access issues or any other questions!

Hi! I clicked on the connection you provided, but I couldn't access it normally, and prompted "Sorry, the server was not able to fulfill your request". Are there any pre-training models still available?