/PIDNet_ade20k

Fine tuning PIDNet on ADE20k dataset

Primary LanguagePython

PIDNet_ade20k

Fine tuning PIDNet on ADE20k dataset

Training

To train the PIDNet on ADE20k dataset, edit the configuration file accordingly specifing the dataset path and other parameters

python tools/train.py --cfg config/pidnet_small_ade20k_trainval.yml

Inference

To infer PIDNet, run the custom.py file present in tools folder.

Results

(The pink color signifies the category of objects that were not part of the training dataset) sample output sample output

Acknowledgement

Reimplementation of https://github.com/XuJiacong/PIDNet

Thanks to the authors of https://arxiv.org/pdf/2206.02066.pdf