Transformer has been considered the dominating neural architecture in NLP and CV, mostly under a supervised setting. Recently, a similar surge of using Transformers has appeared in the domain of reinforcement learning (RL), but it is faced with unique design choices and challenges brought by the nature of RL. However, the evolution of Transformers in RL has not yet been well unraveled. Hence, in this paper, we seek to systematically review motivations and progress on using Transformers in RL, provide a taxonomy on existing works, discuss each sub-field, and summarize future prospects.
It has been witnessed that masked image modeling (MIM) has shown a huge potential in self-supervised learning in the past year. Benefiting from the universal backbone vision transformer, MIM learns self-supervised visual representations through masking a part of patches of the image while attempting to recover the missing pixels. Most previous works mask patches of the image randomly, which underutilizes the semantic information that is beneficial to visual representation learning. On the other hand, due to the large size of the backbone, most previous works have to spend much time on pre-training. In this paper, we propose \textbf{Attention-driven Masking and Throwing Strategy} (AMT), which could solve both problems above. We first leverage the self-attention mechanism to obtain the semantic information of the image during the training process automatically without using any supervised methods. Masking strategy can be guided by that information to mask areas selectively, which is helpful for representation learning. Moreover, a redundant patch throwing strategy is proposed, which makes learning more efficient. As a plug-and-play module for masked image modeling, AMT improves the linear probing accuracy of MAE by $2.9\% \sim 5.9\%$ on CIFAR-10/100, STL-10, Tiny ImageNet, and ImageNet-1K, and obtains an improved performance with respect to fine-tuning accuracy of MAE and SimMIM. Moreover, this design also achieves superior performance on downstream detection and segmentation tasks. Code is available at https://github.com/guijiejie/AMT.
Current scene depth estimation approaches mainly rely on optical sensing, which carries privacy concerns and suffers from estimation ambiguity for distant, shiny, and transparent surfaces/objects. Reconfigurable intelligent surfaces (RISs) provide a path for employing a massive number of antennas using low-cost and energy-efficient architectures. This has the potential for realizing RIS-aided wireless sensing with high spatial resolution. In this paper, we propose to employ RIS-aided wireless sensing systems for scene depth estimation. We develop a comprehensive framework for building accurate depth maps using RIS-aided mmWave sensing systems. In this framework, we propose a new RIS interaction codebook capable of creating a sensing grid of reflected beams that meets the desirable characteristics of efficient scene depth map construction. Using the designed codebook, the received signals are processed to build high-resolution depth maps. Simulation results compare the proposed solution against RGB-based approaches and highlight the promise of adopting RIS-aided mmWave sensing in scene depth perception.
Recent masked image modeling (MIM) has received much attention in self-supervised learning (SSL), which requires the target model to recover the masked part of the input image. Although MIM-based pre-training methods achieve new state-of-the-art performance when transferred to many downstream tasks, the visualizations show that the learned representations are less separable, especially compared to those based on contrastive learning pre-training. This inspires us to think whether the linear separability of MIM pre-trained representation can be further improved, thereby improving the pre-training performance. Since MIM and contrastive learning tend to utilize different data augmentations and training strategies, combining these two pretext tasks is not trivial. In this work, we propose a novel and flexible pre-training framework, named MimCo, which combines MIM and contrastive learning through two-stage pre-training. Specifically, MimCo takes a pre-trained contrastive learning model as the teacher model and is pre-trained with two types of learning targets: patch-level and image-level reconstruction losses. Extensive transfer experiments on downstream tasks demonstrate the superior performance of our MimCo pre-training framework. Taking ViT-S as an example, when using the pre-trained MoCov3-ViT-S as the teacher model, MimCo only needs 100 epochs of pre-training to achieve 82.53% top-1 finetuning accuracy on Imagenet-1K, which outperforms the state-of-the-art self-supervised learning counterparts.
We study the backward compatible problem for person re-identification (Re-ID), which aims to constrain the features of an updated new model to be comparable with the existing features from the old model in galleries. Most of the existing works adopt distillation-based methods, which focus on pushing new features to imitate the distribution of the old ones. However, the distillation-based methods are intrinsically sub-optimal since it forces the new feature space to imitate the inferior old feature space. To address this issue, we propose the Ranking-based Backward Compatible Learning (RBCL), which directly optimizes the ranking metric between new features and old features. Different from previous methods, RBCL only pushes the new features to find best-ranking positions in the old feature space instead of strictly alignment, and is in line with the ultimate goal of backward retrieval. However, the sharp sigmoid function used to make the ranking metric differentiable also incurs the gradient vanish issue, therefore stems the ranking refinement during the later period of training. To address this issue, we propose the Dynamic Gradient Reactivation (DGR), which can reactivate the suppressed gradients by adding dynamic computed constant during forward step. To further help targeting the best-ranking positions, we include the Neighbor Context Agents (NCAs) to approximate the entire old feature space during training. Unlike previous works which only test on the in-domain settings, we make the first attempt to introduce the cross-domain settings (including both supervised and unsupervised), which are more meaningful and difficult. The experimental results on all five settings show that the proposed RBCL outperforms previous state-of-the-art methods by large margins under all settings.
Transformer-based supervised pre-training achieves great performance in person re-identification (ReID). However, due to the domain gap between ImageNet and ReID datasets, it usually needs a larger pre-training dataset (e.g. ImageNet-21K) to boost the performance because of the strong data fitting ability of the transformer. To address this challenge, this work targets to mitigate the gap between the pre-training and ReID datasets from the perspective of data and model structure, respectively. We first investigate self-supervised learning (SSL) methods with Vision Transformer (ViT) pretrained on unlabelled person images (the LUPerson dataset), and empirically find it significantly surpasses ImageNet supervised pre-training models on ReID tasks. To further reduce the domain gap and accelerate the pre-training, the Catastrophic Forgetting Score (CFS) is proposed to evaluate the gap between pre-training and fine-tuning data. Based on CFS, a subset is selected via sampling relevant data close to the down-stream ReID data and filtering irrelevant data from the pre-training dataset. For the model structure, a ReID-specific module named IBN-based convolution stem (ICS) is proposed to bridge the domain gap by learning more invariant features. Extensive experiments have been conducted to fine-tune the pre-training models under supervised learning, unsupervised domain adaptation (UDA), and unsupervised learning (USL) settings. We successfully downscale the LUPerson dataset to 50% with no performance degradation. Finally, we achieve state-of-the-art performance on Market-1501 and MSMT17. For example, our ViT-S/16 achieves 91.3%/89.9%/89.6% mAP accuracy on Market1501 for supervised/UDA/USL ReID. Codes and models will be released to https://github.com/michuanhaohao/TransReID-SSL.
Vision transformers (ViTs) have been an alternative design paradigm to convolutional neural networks (CNNs). However, the training of ViTs is much harder than CNNs, as it is sensitive to the training parameters, such as learning rate, optimizer and warmup epoch. The reasons for training difficulty are empirically analysed in ~\cite{xiao2021early}, and the authors conjecture that the issue lies with the \textit{patchify-stem} of ViT models and propose that early convolutions help transformers see better. In this paper, we further investigate this problem and extend the above conclusion: only early convolutions do not help for stable training, but the scaled ReLU operation in the \textit{convolutional stem} (\textit{conv-stem}) matters. We verify, both theoretically and empirically, that scaled ReLU in \textit{conv-stem} not only improves training stabilization, but also increases the diversity of patch tokens, thus boosting peak performance with a large margin via adding few parameters and flops. In addition, extensive experiments are conducted to demonstrate that previous ViTs are far from being well trained, further showing that ViTs have great potential to be a better substitute of CNNs.
This paper introduces our solution for the Track2 in AI City Challenge 2021 (AICITY21). The Track2 is a vehicle re-identification (ReID) task with both the real-world data and synthetic data. We mainly focus on four points, i.e. training data, unsupervised domain-adaptive (UDA) training, post-processing, model ensembling in this challenge. (1) Both cropping training data and using synthetic data can help the model learn more discriminative features. (2) Since there is a new scenario in the test set that dose not appear in the training set, UDA methods perform well in the challenge. (3) Post-processing techniques including re-ranking, image-to-track retrieval, inter-camera fusion, etc, significantly improve final performance. (4) We ensemble CNN-based models and transformer-based models which provide different representation diversity. With aforementioned techniques, our method finally achieves 0.7445 mAP score, yielding the first place in the competition. Codes are available at https://github.com/michuanhaohao/AICITY2021_Track2_DMT.
Multi-Target Multi-Camera Tracking has a wide range of applications and is the basis for many advanced inferences and predictions. This paper describes our solution to the Track 3 multi-camera vehicle tracking task in 2021 AI City Challenge (AICITY21). This paper proposes a multi-target multi-camera vehicle tracking framework guided by the crossroad zones. The framework includes: (1) Use mature detection and vehicle re-identification models to extract targets and appearance features. (2) Use modified JDETracker (without detection module) to track single-camera vehicles and generate single-camera tracklets. (3) According to the characteristics of the crossroad, the Tracklet Filter Strategy and the Direction Based Temporal Mask are proposed. (4) Propose Sub-clustering in Adjacent Cameras for multi-camera tracklets matching. Through the above techniques, our method obtained an IDF1 score of 0.8095, ranking first on the leaderboard. The code have released: https://github.com/LCFractal/AIC21-MTMC.
In this paper, we explore the Vision Transformer (ViT), a pure transformer-based model, for the object re-identification (ReID) task. With several adaptations, a strong baseline ViT-BoT is constructed with ViT as backbone, which achieves comparable results to convolution neural networks- (CNN-) based frameworks on several ReID benchmarks. Furthermore, two modules are designed in consideration of the specialties of ReID data: (1) It is super natural and simple for Transformer to encode non-visual information such as camera or viewpoint into vector embedding representations. Plugging into these embeddings, ViT holds the ability to eliminate the bias caused by diverse cameras or viewpoints.(2) We design a Jigsaw branch, parallel with the Global branch, to facilitate the training of the model in a two-branch learning framework. In the Jigsaw branch, a jigsaw patch module is designed to learn robust feature representation and help the training of transformer by shuffling the patches. With these novel modules, we propose a pure-transformer framework dubbed as TransReID, which is the first work to use a pure Transformer for ReID research to the best of our knowledge. Experimental results of TransReID are superior promising, which achieve state-of-the-art performance on both person and vehicle ReID benchmarks.