interactive-seg-gui

GUI for tkinter interactive segmentation app.

Installation:

curl -LsSf https://astral.sh/uv/install.sh | sh
# restart your shell

git clone https://github.com/tldr-group/interactive-seg-gui

uv sync --extra cpu
# or for GPU deps
uv sync --extra gpu

If you want to use deep features, you'll need to download the checkpoints:

chmod +x repo/download_chkpoints.sh
./repo/download_chkpoints.sh

Make sure your .env file correctly points to where the models are stored.

Basic use:

1. Add data via Data>Add Image. Features will be computed on upload.
2. Draw labels onto the image to give examples of each class. Use the number keys (1, 2, 3, ...) to change class.
3. Press Train to train a classifier on the labels and apply it to the image.

Advanced tips:

• You can edit the isb_cfg.json file (i.e to change features, post-processing values, ...) whilst the app is running. Press Settings>Reload Config for those changes to be updated on the app. If you want to change the feature stacks after changing the config, press Settings>Re-featurise.
• Hold the Ctrl key and move the mouse to pan the image
• New labels don't overwrite previous labels, you can use this to create good interfacial labels.
• Labels can be saved to disk and loaded again later (Save>Save Labels).
• Comment / uncomment the MODEL_ environment variables in the .env to try different deep feature checkpoints.