fitrialif / Implementation-CVPR2015-CNN-for-ReID

Implementation for CVPR 2015 Paper: "An Improved Deep Learning Architecture for Person Re-Identification".

Implementation-CVPR2015-CNN-for-ReID

---

Notice:

Code for CUHK03 has already been updated. It should be able to be trained now. More detail see the README.

Test Environment:

• Ubuntu 16.04
• Python 2.7.12
• TensorFlow 1.1
• Keras 2.0
• CUDA 8.0
• cudnn 5.1
• NVIDIA GTX 1080Ti

---

Things to do:

• Add some comment to the code for CUHK03.
• Refactor the Data Preparation and Training Routine part of the code.
• Show the Test Result Form.
• Share the trained model to reproduce the test result with Baidu Drive or Google Drive.
• Same to the code for Market-1501.

---

Implementation for CVPR 2015 Paper: "An Improved Deep Learning Architecture for Person Re-Identification".

Paper link

This architechture is implemented based on Keras with Tensorflow backen using Python Programming Language.

Train the model on which dataset?

1. Market-1501

2. CUHK-03

The model structure