search

KaiyangZhou / deep-person-reid

Torchreid: Deep learning person re-identification in PyTorch.
https://kaiyangzhou.github.io/deep-person-reid/
MIT License
4.26k stars 1.14k forks source link

How to use the feature extractor along with Re-Ranking #498

Open 471Q opened 2 years ago

471Q commented 2 years ago

Hello,

I am a bit lost on the documentation of the feature extractor. What I am trying to achieve is I have multiple cameras capturing a person image from different angle, I am getting the person Image I would like to the the extractor mentioned in this repo and would like to re-rank the image from my gallery set.

More specifically I want to know

  1. the feature extractor, what feature is it extracting? is it extracting the complete feature as mentioned on the research paper (global feature, and local feature) [This answer alone might help me build my solution]
  2. can I only use the feature extractor and re-ranking in isolation to achieve what I am trying to achieve

I understand my question might be very broad, please feel free to answer any relevant part of my question or any direction you might think I should take.

Thanks heaps all the intelligent souls

lakshaydulani commented 2 years ago

@471Q you and I are kinda in the same boat. I get lost too in the documentation. I want to know if the person I detected from face recognition in Cam 1 is the same one if he has gone to room 2 (Cam 2).

All I know as of now is that I get a 512 dimension feature vector which I can compare (Euclidean distance) with the other feature vectors.

I just need to track persons in a room at a given time. I need not track their trajectory like deepsort.

SlinkoIgor commented 2 years ago

any news on this question?