search

microsoft / Semantics-Aligned-Representation-Learning-for-Person-Re-identification

This is an implementation of AAAI'20 paper "Semantics-Aligned Representation Learning for Person Re-identification". We leverages dense semantics to address both the spatial misalignment and semantics misalignment challenges in person re-identification.
MIT License
23 stars 13 forks source link

Dataset Preparation #1

Open alakhag opened 4 years ago

alakhag commented 4 years ago

I am trying to train the network on PIT dataset provided, in order to get SAN-PG model. Using this SAN-PG model, I plan to train SAN network.

However, I face issues and solving one of the two may work:

  1. SAN-PG pre-trained model provided cannot be loaded to train SAN, because the architectures are different. Is there any method to load the given SAN-PG model and use it to train the network on CUHK03 data?
  2. The code only works for CUHK03 labelled dataset, and hence, the dataset manager. Is there any method to convert the labelled dataset in the form of CUHK03 data?