BoS
BoS copied to clipboard
[ICLR 2024 Spotlight] Bounding Box Stability against Feature Dropout Reflects Detector Generalization across Environments
Thank you for sharing interesting research result. To run the code, it seems a pre-trained car model on the COCO dataset is needed. I reviewed the provided configuration file for...
The experimental setup provided seems to be for vehicle detection. Has a separate model been trained specifically for pedestrian detection? If so, could you provide the pre-trained model?
Thanks for the great research. According to your paper, "For the RetinaNet+R50 vehicle detector, the best dropout rate ε is 0.15, and the best dropout position setting is adding a...