Support returning bounding box coordinates

YangtaoWANG95 / TokenCut

(CVPR 2022) Pytorch implementation of "Self-supervised transformers for unsupervised object discovery using normalized cut"

MIT License

304 stars 34 forks source link

Support returning bounding box coordinates #4

Closed lextoumbourou closed 2 years ago

lextoumbourou commented 2 years ago

Hi there,

Amazing work with TokenCut! I'm impressed at how well this appears to work on the dataset I'm playing with.

It would be great if there was a convenient method for returning bounding box predictions, instead of just images annotated with the bounding boxes. If that already exists, please point me in the right direction.

XiSHEN0220 commented 2 years ago

Hi,

Thanks for your interest in our work!

You can try these commands which will run TokenCut on a single image.

The bounding box prediction is the first output of the ncut function

Hope this is what you want.

Best,

XiSHEN0220 commented 2 years ago

I close the issue. If you have any other questions, feel free to re-open it.