pse-ecn / pose-sensitive-embedding

Pose Sensitive Embedding for Person Re-Identification (PSE)
112 stars 36 forks source link

Use your pretrained model for feature extraction only #22

Closed cppntn closed 5 years ago

cppntn commented 6 years ago

Hello, thanks for this excellent work. Is there a way to use your pretrained model for features extraction only? I want to test it on frames of a video, so only some detections should be feed into the network for each frame.

Thanks!

ssarfraz commented 6 years ago

you only need to to prepare your data loader for this. look into predictor_preid.py, on changing stuff so ti only writes the features. Also see how the data is fed through one of the datasets.py files, may it would be easy to just follow a similar read.