Large Vision-Language Models (LVLMs) have achieved significant progress in recent years. However, the expensive inference cost limits the realistic deployment of LVLMs. Some works find that visual tokens are redundant and compress tokens to reduce the inference cost. These works identify important non-redundant tokens as target tokens, then prune the remaining tokens (non-target tokens) or merge them into target tokens. However, target token identification faces the token importance-redundancy dilemma. Besides, token merging and pruning face a dilemma between disrupting target token information and losing non-target token information. To solve these problems, we propose a novel visual token compression scheme, named Libra-Merging. In target token identification, Libra-Merging selects the most important tokens from spatially discrete intervals, achieving a more robust token importance-redundancy trade-off than relying on a hyper-parameter. In token compression, when non-target tokens are dissimilar to target tokens, Libra-Merging does not merge them into the target tokens, thus avoiding disrupting target token information. Meanwhile, Libra-Merging condenses these non-target tokens into an information compensation token to prevent losing important non-target token information. Our method can serve as a plug-in for diverse LVLMs, and extensive experimental results demonstrate its effectiveness. The code will be publicly available at https://github.com/longrongyang/Libra-Merging.

The Mixture-of-Experts (MoE) has gained increasing attention in studying Large Vision-Language Models (LVLMs). It uses a sparse model to replace the dense model, achieving comparable performance while activating fewer parameters during inference, thus significantly reducing the inference cost. Existing MoE methods in LVLM encourage different experts to specialize in different tokens, and they usually employ a router to predict the routing of each token. However, the router is not optimized concerning distinct parameter optimization directions generated from tokens within an expert. This may lead to severe interference between tokens within an expert. To address this problem, we propose to use the token-level gradient analysis to Solving Token Gradient Conflict (STGC) in this paper. Specifically, we first use token-level gradients to identify conflicting tokens in experts. After that, we add a regularization loss tailored to encourage conflicting tokens routing from their current experts to other experts, for reducing interference between tokens within an expert. Our method can serve as a plug-in for diverse LVLM methods, and extensive experimental results demonstrate its effectiveness. demonstrate its effectiveness. The code will be publicly available at https://github.com/longrongyang/STGC.

Recently, live streaming platforms have gained immense popularity. Traditional video highlight detection mainly focuses on visual features and utilizes both past and future content for prediction. However, live streaming requires models to infer without future frames and process complex multimodal interactions, including images, audio and text comments. To address these issues, we propose a multimodal transformer that incorporates historical look-back windows. We introduce a novel Modality Temporal Alignment Module to handle the temporal shift of cross-modal signals. Additionally, using existing datasets with limited manual annotations is insufficient for live streaming whose topics are constantly updated and changed. Therefore, we propose a novel Border-aware Pairwise Loss to learn from a large-scale dataset and utilize user implicit feedback as a weak supervision signal. Extensive experiments show our model outperforms various strong baselines on both real-world scenarios and public datasets. And we will release our dataset and code to better assess this topic.

Video understanding is an important task in short video business platforms and it has a wide application in video recommendation and classification. Most of the existing video understanding works only focus on the information that appeared within the video content, including the video frames, audio and text. However, introducing common sense knowledge from the external Knowledge Graph (KG) dataset is essential for video understanding when referring to the content which is less relevant to the video. Owing to the lack of video knowledge graph dataset, the work which integrates video understanding and KG is rare. In this paper, we propose a heterogeneous dataset that contains the multi-modal video entity and fruitful common sense relations. This dataset also provides multiple novel video inference tasks like the Video-Relation-Tag (VRT) and Video-Relation-Video (VRV) tasks. Furthermore, based on this dataset, we propose an end-to-end model that jointly optimizes the video understanding objective with knowledge graph embedding, which can not only better inject factual knowledge into video understanding but also generate effective multi-modal entity embedding for KG. Comprehensive experiments indicate that combining video understanding embedding with factual knowledge benefits the content-based video retrieval performance. Moreover, it also helps the model generate better knowledge graph embedding which outperforms traditional KGE-based methods on VRT and VRV tasks with at least 42.36% and 17.73% improvement in [email protected].

The task of multi-label image classification is to recognize all the object labels presented in an image. Though advancing for years, small objects, and objects with high conditional probability are still the main bottlenecks of previous convolutional neural network (CNN) based models, limited by convolutional kernels' representational capacity. Recent vision transformer networks utilize the self-attention mechanism to extract the feature of pixel granularity. It expresses richer local semantic information, while insufficient for mining global spatial dependence. In this paper, we point out the three crucial problems that CNN-based methods encounter and explore the possibility of conducting specific transformer modules to settle them. We put forward a Multi-label Transformer architecture (MlTr) constructed with windows partitioning, in-window pixel attention, cross-window attention, particularly improving the performance of multi-label image classification tasks. The proposed MlTr shows state-of-the-art results on various prevalent multi-label datasets such as MS-COCO, Pascal-VOC, NUS-WIDE with 88.8%, 95.8%, 65.5% respectively.

In recent years, supervised person re-identification (re-ID) models have received increasing studies. However, these models trained on the source domain always suffer dramatic performance drop when tested on an unseen domain. Existing methods are primary to use pseudo labels to alleviate this problem. One of the most successful approaches predicts neighbors of each unlabeled image and then uses them to train the model. Although the predicted neighbors are credible, they always miss some hard positive samples, which may hinder the model from discovering important discriminative information of the unlabeled domain. In this paper, to complement these low recall neighbor pseudo labels, we propose a joint learning framework to learn better feature embeddings via high precision neighbor pseudo labels and high recall group pseudo labels. The group pseudo labels are generated by transitively merging neighbors of different samples into a group to achieve higher recall. However, the merging operation may cause subgroups in the group due to imperfect neighbor predictions. To utilize these group pseudo labels properly, we propose using a similarity-aggregating loss to mitigate the influence of these subgroups by pulling the input sample towards the most similar embeddings. Extensive experiments on three large-scale datasets demonstrate that our method can achieve state-of-the-art performance under the unsupervised domain adaptation re-ID setting.

As an instance-level recognition problem, re-identification (re-ID) requires models to capture diverse features. However, with continuous training, re-ID models pay more and more attention to the salient areas. As a result, the model may only focus on few small regions with salient representations and ignore other important information. This phenomenon leads to inferior performance, especially when models are evaluated on small inter-identity variation data. In this paper, we propose a novel network, Erasing-Salient Net (ES-Net), to learn comprehensive features by erasing the salient areas in an image. ES-Net proposes a novel method to locate the salient areas by the confidence of objects and erases them efficiently in a training batch. Meanwhile, to mitigate the over-erasing problem, this paper uses a trainable pooling layer P-pooling that generalizes global max and global average pooling. Experiments are conducted on two specific re-identification tasks (i.e., Person re-ID, Vehicle re-ID). Our ES-Net outperforms state-of-the-art methods on three Person re-ID benchmarks and two Vehicle re-ID benchmarks. Specifically, mAP / Rank-1 rate: 88.6% / 95.7% on Market1501, 78.8% / 89.2% on DuckMTMC-reID, 57.3% / 80.9% on MSMT17, 81.9% / 97.0% on Veri-776, respectively. Rank-1 / Rank-5 rate: 83.6% / 96.9% on VehicleID (Small), 79.9% / 93.5% on VehicleID (Medium), 76.9% / 90.7% on VehicleID (Large), respectively. Moreover, the visualized salient areas show human-interpretable visual explanations for the ranking results.