Deep Transfer Learning for Person Re-Identification

By Mengyue Geng, Yaowei Wang, Tao A. Xiang, Yonghong Tian. IEEE International Conference on Multimedia Big Data, 2016

314 citations recorded by the source corpus.

Person re-identification (Re-ID) poses an inevitable challenge to deep learning: how to learn a robust deep model with millions of parameters on a small training set of few or no labels. In this paper, two deep transfer learning methods are proposed to address the training data sparsity problem, respectively from the supervised and unsupervised settings. First, a two-stepped fine-tuning strategy with proxy classifier learning is developed to transfer knowledge from auxiliary datasets. Second, given an unlabelled Re-Iddataset, an unsupervised deep transfer learning model is proposed based on a co-training strategy. Extensive experiments show that the proposed models achieve a good performance of deep Re-ID models.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition, Ocean Engineering

View Paper: https://www.semanticscholar.org/paper/6e0406e765a56f673ea4ff3402d92fd1172826a9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

1. Liang Zheng
2. Yi Yang
3. Zhedong Zheng
4. Qi Tian
5. Shiliang Zhang
6. Wen Gao
7. Tao A. Xiang
8. Shengjin Wang

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Liang Zheng
3. Tao A. Xiang
4. Luc Van Gool
5. Dacheng Tao
6. Xiaogang Wang
7. Qi Tian
8. Ming-Hsuan Yang

In Ocean Engineering

1. Yi Yang
2. Liang Zheng
3. Tao A. Xiang
4. Qi Tian
5. Wanli Ouyang
6. Zhedong Zheng
7. Yifan Sun
8. Shengjin Wang

Citations in the corpus

• Attention-Aware Compositional Network for Person Re-identification (2018, methodology): “Following [16], we build GCN based on the standard GoogleNet [39].”
• Part-Aligned Bilinear Representations for Person Re-identification (2018, methodology): “For 100 test IDs, our method achieves the second best result, following [74].”
• Person Transfer GAN to Bridge Domain Gap for Person Re-identification (2017, methodology): “Some other works [40, 6] combine verification models with classification models to learn descriptors.”
• Improving Person Re-identification by Attribute and Identity Learning (2017, result): “In [45, 8], the combination of verification and classification losses is proven effective, consistent with the findings in [29].”
• Learning With Average Precision: Training Image Retrieval With a Listwise Loss (2019, background): “First, it requires a number of ad hoc heuristics such as pre-training for classification [1, 18], combining several losses [10, 12] and biasing the sampling of image pairs by mining hard or semi-hard negative examples [15, 21, 37, 38].”
• SVDNet for Pedestrian Retrieval (2017, methodology): “So a popular deep learning method for re-ID consists of 1) training a classification deep model on the training set, 2) extracting image descriptors using the fullyconnected (FC) layer for the query and gallery images, and 3) computing similarities based on Euclidean distance before returning the so”
• Furnishing Your Room by What You See: An End-to-End Furniture Set Retrieval Framework with Rich Annotated Benchmark Dataset (2019, methodology): “Some previous works apply item ID or category information as labels to train the feature embedding model[15][9].”
• Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training (2018, background): “2 Transfer [6] 65.”
• Unsupervised Person Re-identification (2017, background): “These works typically deal with the relatively small datasets [16], [17] and to our knowledge, none of these works are integrated into the deep learning framework that has been proven as the state of the art under the large-scale datasets [2], [8], [18], [19].”
• Pose-Normalized Image Generation for Person Re-identification (2017, methodology): “We utilize the two-stepped fine-tuning strategy in [13] to fine-tune ResNet-50-A and ResNet-50-B.”
• A Review of Deep Transfer Learning and Recent Advancements (2022, background): “[27] 2018 Deep Transfer Learning for Person Re-identification Image no Finetuning yes none Identification / security”
• ABD-Net: Attentive but Diverse Person Re-Identification (2019, methodology): “With the ImageNet-pretrained ResNet-50 backbone, we used the two-step transfer learning algorithm [53] to finetune the model.”
• In Defense of the Triplet Loss for Person Re-Identification (2017, methodology): “Especially striking is the jump from 41.5% mAP, obtained by another ResNet-50 model trained with triplets (DTL [8]), to our 69.14% mAP score.”
• Fine-Grained Image Retrieval: the Text/Sketch Input Dilemma (2017, methodology): “We also apply the pairwise dropout strategy [6] for the activation in Siamese branches.”
• AlignedReID: Surpassing Human-Level Performance in Person Re-Identification (2017, background): “Combining softmax loss with metric learning loss to speed up the convergence is also a popular method [10].”
• Pose-Driven Deep Convolutional Model for Person Re-identification (2017, methodology): “More specifically, existing deep learning based person ReID approaches can be summarized into two categories: 1) use Softmax Loss with person ID labels to learn a global representation [1, 10, 53, 46, 42, 61, 13], and 2) first learn local representations using predefined rigid body parts, then fuse ”
• Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in Vitro (2017, methodology): “Recent progress in this field can be attributed to two factors: 1) the availability of large-scale reID datasets [49, 51, 44, 19] and 2) the learned embedding of pedestrians using a CNN [8, 10].”