Weishi Zheng

Peipei Zhu, Xiao Wang, Lin Zhu et al.

Unpaired Image Captioning (UIC) has been developed to learn image descriptions from unaligned vision-language sample pairs. Existing works usually tackle this task using adversarial learning and visual concept reward based on reinforcement learning. However, these existing works were only able to learn limited cross-domain information in vision and language domains, which restrains the captioning performance of UIC. Inspired by the success of Vision-Language Pre-Trained Models (VL-PTMs) in this research, we attempt to infer the cross-domain cue information about a given image from the large VL-PTMs for the UIC task. This research is also motivated by recent successes of prompt learning in many downstream multi-modal tasks, including image-text retrieval and vision question answering. In this work, a semantic prompt is introduced and aggregated with visual features for more accurate caption prediction under the adversarial learning framework. In addition, a metric prompt is designed to select high-quality pseudo image-caption samples obtained from the basic captioning model and refine the model in an iterative manner. Extensive experiments on the COCO and Flickr30K datasets validate the promising captioning ability of the proposed model. We expect that the proposed prompt-based UIC model will stimulate a new line of research for the VL-PTMs based captioning.

Yiqiao Qiu, Yixing Shen, Zhuohao Sun et al.

Continually learning to segment more and more types of image regions is a desired capability for many intelligent systems. However, such continual semantic segmentation suffers from the same catastrophic forgetting issue as in continual classification learning. While multiple knowledge distillation strategies originally for continual classification have been well adapted to continual semantic segmentation, they only consider transferring old knowledge based on the outputs from one or more layers of deep fully convolutional networks. Different from existing solutions, this study proposes to transfer a new type of information relevant to knowledge, i.e. the relationships between elements (Eg. pixels or small local regions) within each image which can capture both within-class and between-class knowledge. The relationship information can be effectively obtained from the self-attention maps in a Transformer-style segmentation model. Considering that pixels belonging to the same class in each image often share similar visual properties, a class-specific region pooling is applied to provide more efficient relationship information for knowledge transfer. Extensive evaluations on multiple public benchmarks support that the proposed self-attention transfer method can further effectively alleviate the catastrophic forgetting issue, and its flexible combination with one or more widely adopted strategies significantly outperforms state-of-the-art solutions.

Qihao Lin, Borui Chen, Yuping Zhou et al.

The contour estimation of transparent fragments is very important for autonomous reassembly, especially in the fields of precision optical instrument repair, cultural relic restoration, and identification of other precious device broken accidents. Different from general intact transparent objects, the contour estimation of transparent fragments face greater challenges due to strict optical properties, irregular shapes and edges. To address this issue, a general transparent fragments contour estimation framework based on visual-tactile fusion is proposed in this paper. First, we construct the transparent fragment dataset named TransFrag27K, which includes a multiscene synthetic data of broken fragments from multiple types of transparent objects, and a scalable synthetic data generation pipeline. Secondly, we propose a visual grasping position detection network named TransFragNet to identify, locate and segment the sampling grasping position. And, we use a two-finger gripper with Gelsight Mini sensors to obtain reconstructed tactile information of the lateral edge of the fragments. By fusing this tactile information with visual cues, a visual-tactile fusion material classifier is proposed. Inspired by the way humans estimate a fragment's contour combining vision and touch, we introduce a general transparent fragment contour estimation framework based on visual-tactile fusion, demonstrates strong performance in real-world validation. Finally, a multi-dimensional similarity metrics based contour matching and reassembly algorithm is proposed, providing a reproducible benchmark for evaluating visual-tactile contour estimation and fragment reassembly. The experimental results demonstrate the validity of the proposed framework. The dataset and codes are available at https://github.com/Keithllin/Transparent-Fragments-Contour-Estimation.

Zhuohao Sun, Yiqiao Qiu, Zhijun Tan et al.

Out-of-distribution (OOD) detection is essential when deploying neural networks in the real world. One main challenge is that neural networks often make overconfident predictions on OOD data. In this study, we propose an effective post-hoc OOD detection method based on a new feature masking strategy and a novel logit smoothing strategy. Feature masking determines the important features at the penultimate layer for each in-distribution (ID) class based on the weights of the ID class in the classifier head and masks the rest features. Logit smoothing computes the cosine similarity between the feature vector of the test sample and the prototype of the predicted ID class at the penultimate layer and uses the similarity as an adaptive temperature factor on the logit to alleviate the network's overconfidence prediction for OOD data. With these strategies, we can reduce feature activation of OOD data and enlarge the gap in OOD score between ID and OOD data. Extensive experiments on multiple standard OOD detection benchmarks demonstrate the effectiveness of our method and its compatibility with existing methods, with new state-of-the-art performance achieved from our method. The source code will be released publicly.

Wei Shi, Wentao Zhang, Weishi Zheng et al.

There is an increasing demand for interpretation of model predictions especially in high-risk applications. Various visualization approaches have been proposed to estimate the part of input which is relevant to a specific model prediction. However, most approaches require model structure and parameter details in order to obtain the visualization results, and in general much effort is required to adapt each approach to multiple types of tasks particularly when model backbone and input format change over tasks. In this study, a simple yet effective visualization framework called PAMI is proposed based on the observation that deep learning models often aggregate features from local regions for model predictions. The basic idea is to mask majority of the input and use the corresponding model output as the relative contribution of the preserved input part to the original model prediction. For each input, since only a set of model outputs are collected and aggregated, PAMI does not require any model detail and can be applied to various prediction tasks with different model backbones and input formats. Extensive experiments on multiple tasks confirm the proposed method performs better than existing visualization approaches in more precisely finding class-specific input regions, and when applied to different model backbones and input formats. The source code will be released publicly.

Yu Zhou, Xiaoyan Yang, Bojia Zi et al.

This paper introduces Point2Insert, a sparse-point-based framework for flexible and user-friendly object insertion in videos, motivated by the growing popularity of accurate, low-effort object placement. Existing approaches face two major challenges: mask-based insertion methods require labor-intensive mask annotations, while instruction-based methods struggle to place objects at precise locations. Point2Insert addresses these issues by requiring only a small number of sparse points instead of dense masks, eliminating the need for tedious mask drawing. Specifically, it supports both positive and negative points to indicate regions that are suitable or unsuitable for insertion, enabling fine-grained spatial control over object locations. The training of Point2Insert consists of two stages. In Stage 1, we train an insertion model that generates objects in given regions conditioned on either sparse-point prompts or a binary mask. In Stage 2, we further train the model on paired videos synthesized by an object removal model, adapting it to video insertion. Moreover, motivated by the higher insertion success rate of mask-guided editing, we leverage a mask-guided insertion model as a teacher to distill reliable insertion behavior into the point-guided model. Extensive experiments demonstrate that Point2Insert consistently outperforms strong baselines and even surpasses models with $\times$10 more parameters.

Yihang Guo, Tianyuan Yu, Liang Bai et al.

Multi-task learning (MTL) aims to build general-purpose vision systems by training a single network to perform multiple tasks jointly. While promising, its potential is often hindered by "unbalanced optimization", where task interference leads to subpar performance compared to single-task models. To facilitate research in MTL, this paper presents a systematic experimental analysis to dissect the factors contributing to this persistent problem. Our investigation confirms that the performance of existing optimization methods varies inconsistently across datasets, and advanced architectures still rely on costly grid-searched loss weights. Furthermore, we show that while powerful Vision Foundation Models (VFMs) provide strong initialization, they do not inherently resolve the optimization imbalance, and merely increasing data quantity offers limited benefits. A crucial finding emerges from our analysis: a strong correlation exists between the optimization imbalance and the norm of task-specific gradients. We demonstrate that this insight is directly applicable, showing that a straightforward strategy of scaling task losses according to their gradient norms can achieve performance comparable to that of an extensive and computationally expensive grid search. Our comprehensive analysis suggests that understanding and controlling gradient dynamics is a more direct path to stable MTL than developing increasingly complex methods.

Xuchen Xie, Yiqiao Qiu, Run Lin et al.

This study focuses on incremental learning for image classification, exploring how to reduce catastrophic forgetting of all learned knowledge when access to old data is restricted due to memory or privacy constraints. The challenge of incremental learning lies in achieving an optimal balance between plasticity, the ability to learn new knowledge, and stability, the ability to retain old knowledge. Based on whether the task identifier (task-ID) of an image can be obtained during the test stage, incremental learning for image classifcation is divided into two main paradigms, which are task incremental learning (TIL) and class incremental learning (CIL). The TIL paradigm has access to the task-ID, allowing it to use multiple task-specific classification heads selected based on the task-ID. Consequently, in CIL, where the task-ID is unavailable, TIL methods must predict the task-ID to extend their application to the CIL paradigm. Our previous method for TIL adds task-specific batch normalization and classification heads incrementally. This work extends the method by predicting task-ID through an "unknown" class added to each classification head. The head with the lowest "unknown" probability is selected, enabling task-ID prediction and making the method applicable to CIL. The task-specific batch normalization (BN) modules effectively adjust the distribution of output feature maps across different tasks, enhancing the model's plasticity.Moreover, since BN has much fewer parameters compared to convolutional kernels, by only modifying the BN layers as new tasks arrive, the model can effectively manage parameter growth while ensuring stability across tasks. The innovation of this study lies in the first-time introduction of task-specific BN into CIL and verifying the feasibility of extending TIL methods to CIL through task-ID prediction with state-of-the-art performance on multiple datasets.

Changcai Li, Wenwei Lin, Zuoxun Hou et al.

In this work, we explore the technical feasibility of implementing end-to-end 3D object detection (3DOD) with surround-view fisheye camera system. Specifically, we first investigate the performance drop incurred when transferring classic pinhole-based 3D object detectors to fisheye imagery. To mitigate this, we then develop two methods that incorporate the unique geometry of fisheye images into mainstream detection frameworks: one based on the bird's-eye-view (BEV) paradigm, named FisheyeBEVDet, and the other on the query-based paradigm, named FisheyePETR. Both methods adopt spherical spatial representations to effectively capture fisheye geometry. In light of the lack of dedicated evaluation benchmarks, we release Fisheye3DOD, a new open dataset synthesized using CARLA and featuring both standard pinhole and fisheye camera arrays. Experiments on Fisheye3DOD show that our fisheye-compatible modeling improves accuracy by up to 6.2% over baseline methods.

Zhihui Zhu, Xinyang Jiang, Feng Zheng et al.

Although great progress in supervised person re-identification (Re-ID) has been made recently, due to the viewpoint variation of a person, Re-ID remains a massive visual challenge. Most existing viewpoint-based person Re-ID methods project images from each viewpoint into separated and unrelated sub-feature spaces. They only model the identity-level distribution inside an individual viewpoint but ignore the underlying relationship between different viewpoints. To address this problem, we propose a novel approach, called \textit{Viewpoint-Aware Loss with Angular Regularization }(\textbf{VA-reID}). Instead of one subspace for each viewpoint, our method projects the feature from different viewpoints into a unified hypersphere and effectively models the feature distribution on both the identity-level and the viewpoint-level. In addition, rather than modeling different viewpoints as hard labels used for conventional viewpoint classification, we introduce viewpoint-aware adaptive label smoothing regularization (VALSR) that assigns the adaptive soft label to feature representation. VALSR can effectively solve the ambiguity of the viewpoint cluster label assignment. Extensive experiments on the Market1501 and DukeMTMC-reID datasets demonstrated that our method outperforms the state-of-the-art supervised Re-ID methods.

Xinyang Jiang, Yifei Gong, Xiaowei Guo et al.

Recently, the research interest of person re-identification (ReID) has gradually turned to video-based methods, which acquire a person representation by aggregating frame features of an entire video. However, existing video-based ReID methods do not consider the semantic difference brought by the outputs of different network stages, which potentially compromises the information richness of the person features. Furthermore, traditional methods ignore important relationship among frames, which causes information redundancy in fusion along the time axis. To address these issues, we propose a novel general temporal fusion framework to aggregate frame features on both semantic aspect and time aspect. As for the semantic aspect, a multi-stage fusion network is explored to fuse richer frame features at multiple semantic levels, which can effectively reduce the information loss caused by the traditional single-stage fusion. While, for the time axis, the existing intra-frame attention method is improved by adding a novel inter-frame attention module, which effectively reduces the information redundancy in temporal fusion by taking the relationship among frames into consideration. The experimental results show that our approach can effectively improve the video-based re-identification accuracy, achieving the state-of-the-art performance.

Hui Li, Meng Yang, Zhihui Lai et al.

Deep part-based methods in recent literature have revealed the great potential of learning local part-level representation for pedestrian image in the task of person re-identification. However, global features that capture discriminative holistic information of human body are usually ignored or not well exploited. This motivates us to investigate joint learning global and local features from pedestrian images. Specifically, in this work, we propose a novel framework termed tree branch network (TBN) for person re-identification. Given a pedestrain image, the feature maps generated by the backbone CNN, are partitioned recursively into several pieces, each of which is followed by a bottleneck structure that learns finer-grained features for each level in the hierarchical tree-like framework. In this way, representations are learned in a coarse-to-fine manner and finally assembled to produce more discriminative image descriptions. Experimental results demonstrate the effectiveness of the global and local feature learning method in the proposed TBN framework. We also show significant improvement in performance over state-of-the-art methods on three public benchmarks: Market-1501, CUHK-03 and DukeMTMC.

Minghan Li, Tanli Zuo, Ruicheng Li et al.

Knowledge distillation is an effective technique that transfers knowledge from a large teacher model to a shallow student. However, just like massive classification, large scale knowledge distillation also imposes heavy computational costs on training models of deep neural networks, as the softmax activations at the last layer involve computing probabilities over numerous classes. In this work, we apply the idea of importance sampling which is often used in Neural Machine Translation on large scale knowledge distillation. We present a method called dynamic importance sampling, where ranked classes are sampled from a dynamic distribution derived from the interaction between the teacher and student in full distillation. We highlight the utility of our proposal prior which helps the student capture the main information in the loss function. Our approach manages to reduce the computational cost at training time while maintaining the competitive performance on CIFAR-100 and Market-1501 person re-identification datasets.

Yongyi Tang, Lin Ma, Wei Liu et al.

Human motion prediction aims at generating future frames of human motion based on an observed sequence of skeletons. Recent methods employ the latest hidden states of a recurrent neural network (RNN) to encode the historical skeletons, which can only address short-term prediction. In this work, we propose a motion context modeling by summarizing the historical human motion with respect to the current prediction. A modified highway unit (MHU) is proposed for efficiently eliminating motionless joints and estimating next pose given the motion context. Furthermore, we enhance the motion dynamic by minimizing the gram matrix loss for long-term motion prediction. Experimental results show that the proposed model can promisingly forecast the human future movements, which yields superior performances over related state-of-the-art approaches. Moreover, specifying the motion context with the activity labels enables our model to perform human motion transfer.

Hailin Shi, Yang Yang, Xiangyu Zhu et al.

Person re-identification is challenging due to the large variations of pose, illumination, occlusion and camera view. Owing to these variations, the pedestrian data is distributed as highly-curved manifolds in the feature space, despite the current convolutional neural networks (CNN)'s capability of feature extraction. However, the distribution is unknown, so it is difficult to use the geodesic distance when comparing two samples. In practice, the current deep embedding methods use the Euclidean distance for the training and test. On the other hand, the manifold learning methods suggest to use the Euclidean distance in the local range, combining with the graphical relationship between samples, for approximating the geodesic distance. From this point of view, selecting suitable positive i.e. intra-class) training samples within a local range is critical for training the CNN embedding, especially when the data has large intra-class variations. In this paper, we propose a novel moderate positive sample mining method to train robust CNN for person re-identification, dealing with the problem of large variation. In addition, we improve the learning by a metric weight constraint, so that the learned metric has a better generalization ability. Experiments show that these two strategies are effective in learning robust deep metrics for person re-identification, and accordingly our deep model significantly outperforms the state-of-the-art methods on several benchmarks of person re-identification. Therefore, the study presented in this paper may be useful in inspiring new designs of deep models for person re-identification.