Multi-Level Feature Learning for Person Re-Identification
Master's thesis on deep discriminative feature learning for person re-identification.
Project Snapshot
| Signal | Details |
|---|---|
| Timeline | Jun 2018 to Apr 2019 |
| Organization | Xiamen University |
| Focus | Deep Discriminative Features Learning with Multi-Level Network for Person Re-Identification |
| Stack | Custom implementation and project-specific tooling |
Thesis Title:
A multi-level network is proposed to extract the deep discriminative features for person re-identification task
Contributions
- Proposed a multi-level network for extracting discriminative person ReID features.
- Applied batch-hard triplet loss at multiple network levels for stronger metric learning.
- Fused features from different levels for the final classification task.