The proposed architecture is implemented using the PyTorch framework (1.9.0+cu111) with a single GeForce RTX 3090 GPU of 24 GB memory.
We have used the following datasets:
All the dataset follows an 80:10:10 split for training, validation and testing, except for the Kvasir-SEG, where the dataset is split into training and testing.
You can download the weight file from the following links:
@inproceedings{tomar2022tganet title={TGANet: Text-guided attention for improved polyp segmentation}, author={Tomar, Nikhil Kumar and Jha, Debesh and Bagci, Ulas and Ali, Sharib}, booktitle={arXiv preprint arXiv:2205.04280}, year={2022} }
The source code is free for research and education use only. Any comercial use should receive a formal permission from the first author.
Please contact [email protected] for any further questions.