Diversity-Achieving Slow-DropBlock Network for Person Re-Identification
This work addresses feature diversity in person re-identification, an incremental improvement for computer vision applications.
The paper tackles the challenge of learning diverse features in person re-identification by proposing a network that moves dropping operations from intermediate features to input images, using a double-batch-split co-training approach to handle convergence issues. It achieves superior performance over the baseline BDB method on datasets like Market-1501, DukeMTMC-reID, and CUHK03, with further gains from more dropping branches.
A big challenge of person re-identification (Re-ID) using a multi-branch network architecture is to learn diverse features from the ID-labeled dataset. The 2-branch Batch DropBlock (BDB) network was recently proposed for achieving diversity between the global branch and the feature-dropping branch. In this paper, we propose to move the dropping operation from the intermediate feature layer towards the input (image dropping). Since it may drop a large portion of input images, this makes the training hard to converge. Hence, we propose a novel double-batch-split co-training approach for remedying this problem. In particular, we show that the feature diversity can be well achieved with the use of multiple dropping branches by setting individual dropping ratio for each branch. Empirical evidence demonstrates that the proposed method performs superior to BDB on popular person Re-ID datasets, including Market-1501, DukeMTMC-reID and CUHK03 and the use of more dropping branches can further boost the performance.