Jiawei Li

Kaicheng Yang, Jiankang Deng, Xiang An et al.

Contrastive Language-Image Pre-training (CLIP) has significantly boosted the performance of various vision-language tasks by scaling up the dataset with image-text pairs collected from the web. However, the presence of intrinsic noise and unmatched image-text pairs in web data can potentially affect the performance of representation learning. To address this issue, we first utilize the OFA model to generate synthetic captions that focus on the image content. The generated captions contain complementary information that is beneficial for pre-training. Then, we propose an Adaptive Language-Image Pre-training (ALIP), a bi-path model that integrates supervision from both raw text and synthetic caption. As the core components of ALIP, the Language Consistency Gate (LCG) and Description Consistency Gate (DCG) dynamically adjust the weights of samples and image-text/caption pairs during the training process. Meanwhile, the adaptive contrastive loss can effectively reduce the impact of noise data and enhances the efficiency of pre-training data. We validate ALIP with experiments on different scales of models and pre-training datasets. Experiments results show that ALIP achieves state-of-the-art performance on multiple downstream tasks including zero-shot image-text retrieval and linear probe. To facilitate future research, the code and pre-trained models are released at https://github.com/deepglint/ALIP.

Enze Xie, Lewei Yao, Han Shi et al.

Diffusion models have proven to be highly effective in generating high-quality images. However, adapting large pre-trained diffusion models to new domains remains an open challenge, which is critical for real-world applications. This paper proposes DiffFit, a parameter-efficient strategy to fine-tune large pre-trained diffusion models that enable fast adaptation to new domains. DiffFit is embarrassingly simple that only fine-tunes the bias term and newly-added scaling factors in specific layers, yet resulting in significant training speed-up and reduced model storage costs. Compared with full fine-tuning, DiffFit achieves 2$\times$ training speed-up and only needs to store approximately 0.12\% of the total model parameters. Intuitive theoretical analysis has been provided to justify the efficacy of scaling factors on fast adaptation. On 8 downstream datasets, DiffFit achieves superior or competitive performances compared to the full fine-tuning while being more efficient. Remarkably, we show that DiffFit can adapt a pre-trained low-resolution generative model to a high-resolution one by adding minimal cost. Among diffusion-based methods, DiffFit sets a new state-of-the-art FID of 3.02 on ImageNet 512$\times$512 benchmark by fine-tuning only 25 epochs from a public pre-trained ImageNet 256$\times$256 checkpoint while being 30$\times$ more training efficient than the closest competitor.

Tianyang Hu, Fei Chen, Haonan Wang et al.

In generative modeling, numerous successful approaches leverage a low-dimensional latent space, e.g., Stable Diffusion models the latent space induced by an encoder and generates images through a paired decoder. Although the selection of the latent space is empirically pivotal, determining the optimal choice and the process of identifying it remain unclear. In this study, we aim to shed light on this under-explored topic by rethinking the latent space from the perspective of model complexity. Our investigation starts with the classic generative adversarial networks (GANs). Inspired by the GAN training objective, we propose a novel "distance" between the latent and data distributions, whose minimization coincides with that of the generator complexity. The minimizer of this distance is characterized as the optimal data-dependent latent that most effectively capitalizes on the generator's capacity. Then, we consider parameterizing such a latent distribution by an encoder network and propose a two-stage training strategy called Decoupled Autoencoder (DAE), where the encoder is only updated in the first stage with an auxiliary decoder and then frozen in the second stage while the actual decoder is being trained. DAE can improve the latent distribution and as a result, improve the generative performance. Our theoretical analyses are corroborated by comprehensive experiments on various models such as VQGAN and Diffusion Transformer, where our modifications yield significant improvements in sample quality with decreased model complexity.

Yujun Huang, Bin Chen, Shiyu Qin et al.

Beyond achieving higher compression efficiency over classical image compression codecs, deep image compression is expected to be improved with additional side information, e.g., another image from a different perspective of the same scene. To better utilize the side information under the distributed compression scenario, the existing method (Ayzik and Avidan 2020) only implements patch matching at the image domain to solve the parallax problem caused by the difference in viewing points. However, the patch matching at the image domain is not robust to the variance of scale, shape, and illumination caused by the different viewing angles, and can not make full use of the rich texture information of the side information image. To resolve this issue, we propose Multi-Scale Feature Domain Patch Matching (MSFDPM) to fully utilizes side information at the decoder of the distributed image compression model. Specifically, MSFDPM consists of a side information feature extractor, a multi-scale feature domain patch matching module, and a multi-scale feature fusion network. Furthermore, we reuse inter-patch correlation from the shallow layer to accelerate the patch matching of the deep layer. Finally, we nd that our patch matching in a multi-scale feature domain further improves compression rate by about 20% compared with the patch matching method at image domain (Ayzik and Avidan 2020).

Jiawei Li, Chengye Yang, Yaochen Zhang et al.

This study addresses construction site hazard identification by proposing a retrieval-augmented framework that enhances large language models (LLMs) without requiring fine-tuning. Current LLM-based approaches face limitations: image-text matching struggles with complex hazards, while instruction tuning lacks generalization and is resource-intensive. Our method dynamically integrates external knowledge and retrieved similar cases via prompt tuning, overcoming LLMs' limitations in domain knowledge and feature correlation. The framework comprises a case database, an image retrieval module, and an LLM-based reasoning module. Evaluated on real-site data, our approach boosted GLM-4V's accuracy to 50%, a 35.49% improvement over baselines, with consistent gains across hazard types. Ablation studies validated the effectiveness of our image retrieval strategy, showing the superiority of our LPIPS- and CLIP-based method. The proposed technique significantly improves identification accuracy and contextual understanding, demonstrating strong generalization and offering a practical path for intelligent safety risk detection in construction.

Ruxin Ding, Jianfeng Ren, Heng Yu et al.

Spatial-temporal local binary pattern (STLBP) has been widely used in dynamic texture recognition. STLBP often encounters the high-dimension problem as its dimension increases exponentially, so that STLBP could only utilize a small neighborhood. To tackle this problem, we propose a method for dynamic texture recognition using PDV hashing and dictionary learning on multi-scale volume local binary pattern (PHD-MVLBP). Instead of forming very high-dimensional LBP histogram features, it first uses hash functions to map the pixel difference vectors (PDVs) to binary vectors, then forms a dictionary using the derived binary vector, and encodes them using the derived dictionary. In such a way, the PDVs are mapped to feature vectors of the size of dictionary, instead of LBP histograms of very high dimension. Such an encoding scheme could extract the discriminant information from videos in a much larger neighborhood effectively. The experimental results on two widely-used dynamic textures datasets, DynTex++ and UCLA, show the superiority performance of the proposed approach over the state-of-the-art methods.

Hao Fang, Daoxin Zhang, Yi Zhang et al.

Instance segmentation can detect where the objects are in an image, but hard to understand the relationship between them. We pay attention to a typical relationship, relative saliency. A closely related task, salient object detection, predicts a binary map highlighting a visually salient region while hard to distinguish multiple objects. Directly combining two tasks by post-processing also leads to poor performance. There is a lack of research on relative saliency at present, limiting the practical applications such as content-aware image cropping, video summary, and image labeling. In this paper, we study the Salient Object Ranking (SOR) task, which manages to assign a ranking order of each detected object according to its visual saliency. We propose the first end-to-end framework of the SOR task and solve it in a multi-task learning fashion. The framework handles instance segmentation and salient object ranking simultaneously. In this framework, the SOR branch is independent and flexible to cooperate with different detection methods, so that easy to use as a plugin. We also introduce a Position-Preserved Attention (PPA) module tailored for the SOR branch. It consists of the position embedding stage and feature interaction stage. Considering the importance of position in saliency comparison, we preserve absolute coordinates of objects in ROI pooling operation and then fuse positional information with semantic features in the first stage. In the feature interaction stage, we apply the attention mechanism to obtain proposals' contextualized representations to predict their relative ranking orders. Extensive experiments have been conducted on the ASR dataset. Without bells and whistles, our proposed method outperforms the former state-of-the-art method significantly. The code will be released publicly available.

Fengwei Zhou, Jiawei Li, Chuanlong Xie et al.

Automated data augmentation has shown superior performance in image recognition. Existing works search for dataset-level augmentation policies without considering individual sample variations, which are likely to be sub-optimal. On the other hand, learning different policies for different samples naively could greatly increase the computing cost. In this paper, we learn a sample-aware data augmentation policy efficiently by formulating it as a sample reweighting problem. Specifically, an augmentation policy network takes a transformation and the corresponding augmented image as inputs, and outputs a weight to adjust the augmented image loss computed by a task network. At training stage, the task network minimizes the weighted losses of augmented training images, while the policy network minimizes the loss of the task network on a validation set via meta-learning. We theoretically prove the convergence of the training procedure and further derive the exact convergence rate. Superior performance is achieved on widely-used benchmarks including CIFAR-10/100, Omniglot, and ImageNet.

Yuting Xiao, Yanyu Xu, Ziming Zhong et al.

Almost all existing amodal segmentation methods make the inferences of occluded regions by using features corresponding to the whole image. This is against the human's amodal perception, where human uses the visible part and the shape prior knowledge of the target to infer the occluded region. To mimic the behavior of human and solve the ambiguity in the learning, we propose a framework, it firstly estimates a coarse visible mask and a coarse amodal mask. Then based on the coarse prediction, our model infers the amodal mask by concentrating on the visible region and utilizing the shape prior in the memory. In this way, features corresponding to background and occlusion can be suppressed for amodal mask estimation. Consequently, the amodal mask would not be affected by what the occlusion is given the same visible regions. The leverage of shape prior makes the amodal mask estimation more robust and reasonable. Our proposed model is evaluated on three datasets. Experiments show that our proposed model outperforms existing state-of-the-art methods. The visualization of shape prior indicates that the category-specific feature in the codebook has certain interpretability.

Xingchun Xiang, Qingtao Tang, Huaixuan Zhang et al.

Regression tree (RT) has been widely used in machine learning and data mining community. Given a target data for prediction, a regression tree is first constructed based on a training dataset before making prediction for each leaf node. In practice, the performance of RT relies heavily on the local mean of samples from an individual node during the tree construction/prediction stage, while neglecting the global information from different nodes, which also plays an important role. To address this issue, we propose a novel regression tree, named James-Stein Regression Tree (JSRT) by considering global information from different nodes. Specifically, we incorporate the global mean information based on James-Stein estimator from different nodes during the construction/predicton stage. Besides, we analyze the generalization error of our method under the mean square error (MSE) metric. Extensive experiments on public benchmark datasets verify the effectiveness and efficiency of our method, and demonstrate the superiority of our method over other RT prediction methods.

Jiawei Li, Jae Chul Koh, Won-Sook Lee

Image interpolation in medical area is of high importance as most 3D biomedical volume images are sampled where the distance between consecutive slices significantly greater than the in-plane pixel size due to radiation dose or scanning time. Image interpolation creates a number of new slices between known slices in order to obtain an isotropic volume image. The results can be used for the higher quality of 3D reconstruction and visualization of human body structures. Semantic interpolation on the manifold has been proved to be very useful for smoothing image interpolation. Nevertheless, all previous methods focused on low-resolution image interpolation, and most of them work poorly on high-resolution image. We propose a novel network, High Resolution Interpolation Network (HRINet), aiming at producing high-resolution CT image interpolations. We combine the idea of ACAI and GANs, and propose a novel idea of alternative supervision method by applying supervised and unsupervised training alternatively to raise the accuracy of human organ structures in CT while keeping high quality. We compare an MSE based and a perceptual based loss optimizing methods for high quality interpolation, and show the tradeoff between the structural correctness and sharpness. Our experiments show the great improvement on 256 2 and 5122 images quantitatively and qualitatively.

Mang Ye, Andy J Ma, Liang Zheng et al.

Label estimation is an important component in an unsupervised person re-identification (re-ID) system. This paper focuses on cross-camera label estimation, which can be subsequently used in feature learning to learn robust re-ID models. Specifically, we propose to construct a graph for samples in each camera, and then graph matching scheme is introduced for cross-camera labeling association. While labels directly output from existing graph matching methods may be noisy and inaccurate due to significant cross-camera variations, this paper proposes a dynamic graph matching (DGM) method. DGM iteratively updates the image graph and the label estimation process by learning a better feature space with intermediate estimated labels. DGM is advantageous in two aspects: 1) the accuracy of estimated labels is improved significantly with the iterations; 2) DGM is robust to noisy initial training data. Extensive experiments conducted on three benchmarks including the large-scale MARS dataset show that DGM yields competitive performance to fully supervised baselines, and outperforms competing unsupervised learning methods.