Pure PyTorch implementation of Superpixel Sampling Networks that does not require CUDA code compilation. Some minor modifications were also made.
Forked from: https://github.com/perrying/ssn-pytorch
paper: https://arxiv.org/abs/1807.10174
original code: https://github.com/NVlabs/ssn_superpixels
- PyTorch >= 1.4 (This fork was tested only with 1.12)
- scikit-image
- matplotlib
SSN_pix
python inference.py --image /path/to/image
SSN_deep
python inference.py --image /path/to/image --weight /path/to/pretrained_weight
python train.py --root /path/to/BSDS500
SSN_pix
SSN_deep