Adversarially Occluded Samples for Person Re-identification
Citations Over TimeTop 1% of 2018 papers
Abstract
Person re-identification (ReID) is the task of retrieving particular persons across different cameras. Despite its great progress in recent years, it is still confronted with challenges like pose variation, occlusion, and similar appearance among different persons. The large gap between training and testing performance with existing models implies the insufficiency of generalization. Considering this fact, we propose to augment the variation of training data by introducing Adversarially Occluded Samples. These special samples are both a) meaningful in that they resemble real-scene occlusions, and b) effective in that they are tough for the original model and thus provide the momentum to jump out of local optimum. We mine these samples based on a trained ReID model and with the help of network visualization techniques. Extensive experiments show that the proposed samples help the model discover new discriminative clues on the body and generalize much better at test time. Our strategy makes significant improvement over strong baselines on three large-scale ReID datasets, Market1501, CUHK03 and DukeMTMC-reID.
Related Papers
- → Searching and Learning Discriminative Regions for Fine-Grained Image Retrieval and Classification(2021)6 cited
- Stable Discriminative Dictionary Learning Via Discriminative Deviation(2012)
- → Accurate And Fast Fine-Grained Image Classification via Discriminative Learning(2019)3 cited
- Classification Control of Project Variation(2007)