For example, I have a faster-rcnn model pre-trained on coco dataset and I am fine-tuning on downstream dataset. To analyse the features of the detector, I want to extract the embeddings from the neck (i.e., FPN), but I am not sure how to extract the feature embeddings. Can you help me understand and implement it?
For example, I have a faster-rcnn model pre-trained on coco dataset and I am fine-tuning on downstream dataset. To analyse the features of the detector, I want to extract the embeddings from the neck (i.e., FPN), but I am not sure how to extract the feature embeddings. Can you help me understand and implement it?