Get additional output results (feature maps) from the existing object detector models.

[WoojuLee24]

Hello, I am planning to get feature maps of TensorRT object detector models. TensorRT object detector models include Faster R-CNN and Mask R-CNN model powered by TensorRT engine. But provided demo code output only bboxes, labels, and masks. I want to output feature maps of object detector models.

detector = Detector(model_path=args.model_path, device_name, device_id=0) bboxes, labels, masks = detector(img) -> detector = Detector(model_path=args.model_path, device_name, device_id=0) bboxes, labels, masks, feature_maps = detector(img)

The Detector class seem to be compiled before implementing python demo code. How can I output feature maps of existing object detector models? I would appreciate if you give any help. Thank you.

[tpoisonooo]

@lvhan028

[lvhan028]

Do you mean to get the output of model neck?

[WoojuLee24]

@lvhan028 Thank you for replying. Yes, I want to get the outputs of model neck.