| title | Atmoscope |
|---|---|
| emoji | ☁️ |
| colorFrom | blue |
| colorTo | indigo |
| sdk | docker |
| pinned | false |
| short_description | Cloud type classification with ConvNeXt and NASA GAZE data |
Atmoscope is an end-to-end ML project focused on cloud type (genus) classification from phone pictures. This repo is for the live model deployed on Hugging Face!
This is a deployment of a multi-label cloud classifier trained on the NASA GAZE dataset. The model itself is ConvNeXt with a linear layer, and was fine tuned from ImageNet pretrained weights. Parts of the process are documented on my blog. The remote repo is on GitHub and Hugging Face.