Hyperparameters for image feature extraction

Hello, thanks for open-sourcing the amazing work!
Do you mind sharing more experiment details for image feature extraction?
For example, what are the score threshold, NMS threshold, and the number of detections to keep?
For every image, do you always generate the same amount of detections so that the pooled ROI feature dimension will be the same for all images? Thank you again for your time!

No file, test, or entry point is named. Start by locating the image feature extraction setup and its experiment configuration, then verify the score threshold, NMS threshold, detection limit, and whether detection counts vary by image. Done means these details are documented clearly, including the resulting pooled ROI feature dimensions.

Written by the indexing model from the issue text.