Open 2462954048 opened 2 months ago
Since evidential deep learning only requires treating the logit output of the previous classification head as evidence, you should be able to run the process by simply modifying the faster_rcnn.py file in the EOD/EOD/modeling/roi_heads. To reduce training time, you can directly fine-tune using the weights provided in (https://huggingface.co/natewang021/eod/tree/main).
Based on the instructions you provided, I found it difficult to reproduce the project. After setting up your EOD project and installing detectron2 and opendet in the project, I am not sure how to proceed next. Can you provide more clear step-by-step instructions for Basic Setup and Usage?