Novel view synthesis has demonstrated impressive progress recently, with 3D Gaussian splatting (3DGS) offering efficient training time and photorealistic real-time rendering. However, reliance on Cartesian coordinates limits 3DGS’s performance on distant objects, which is important for reconstructing unbounded outdoor environments. We found that, despite its ultimate simplicity, using homogeneous coordinates, a concept on the projective geometry, for the 3DGS pipeline remarkably improves the rendering accuracies of distant objects. We therefore propose Homogeneous Gaussian Splatting (HoGS) incorporating homogeneous coordinates into the 3DGS framework, providing a unified representation for enhancing near and distant objects. HoGS effectively manages both expansive spatial positions and scales particularly in outdoor unbounded environments by adopting projective geometry principles. Experiments show that HoGS significantly enhances accuracy in reconstructing distant objects while maintaining high-quality rendering of nearby objects, along with fast training speed and real-time rendering capability. Our implementations are available on our project page https://kh129.github.io/hogs/.

Long-form video understanding with Large Vision Language Models is challenged by the need to analyze temporally dispersed yet spatially concentrated key moments within limited context windows. In this work, we introduce VideoMindPalace, a new framework inspired by the "Mind Palace", which organizes critical video moments into a topologically structured semantic graph. VideoMindPalace organizes key information through (i) hand-object tracking and interaction, (ii) clustered activity zones representing specific areas of recurring activities, and (iii) environment layout mapping, allowing natural language parsing by LLMs to provide grounded insights on spatio-temporal and 3D context. In addition, we propose the Video Mind-Palace Benchmark (VMB), to assess human-like reasoning, including spatial localization, temporal reasoning, and layout-aware sequential understanding. Evaluated on VMB and established video QA datasets, including EgoSchema, NExT-QA, IntentQA, and the Active Memories Benchmark, VideoMindPalace demonstrates notable gains in spatiotemporal coherence and human-aligned reasoning, advancing long-form video analysis capabilities in VLMs. Project website: https://oodbag.github.io/VMP/.

Text-guided image manipulation has experienced notable advancement in recent years. In order to mitigate linguistic ambiguity, few-shot learning with visual examples has been applied for instructions that are underrepresented in the training set, or difficult to describe purely in language. However, learning from visual prompts requires strong reasoning capability, which diffusion models are struggling with. To address this issue, we introduce a novel multi-modal autoregressive model, dubbed InstaManip, that can instantly learn a new image manipulation operation from textual and visual guidance via in-context learning, and apply it to new query images. Specifically, we propose an innovative group self-attention mechanism to break down the in-context learning process into two separate stages – learning and applying, which simplifies the complex problem into two easier tasks. We also introduce a relation regularization method to further disentangle image transformation features from irrelevant contents in exemplar images. Extensive experiments suggest that our method surpasses previous few-shot image manipulation models by a notable margin (≥19% in human evaluation). We also find our model can be further boosted by increasing the number or diversity of exemplar images. Please check out our project page (https://bolinlai.github.io/projects/InstaManip/).

Video action recognition involves interpreting both global context and specific details to accurately identify actions. While previous models are effective at capturing spatiotemporal features, they often lack a focused representation of key action details. To address this, we introduce FocusVideo, a framework designed for refining video action recognition through integrated global and local feature learning. Inspired by human visual cognition theory, our approach balances the focus on both broad contextual changes and action-specific details, minimizing the influence of irrelevant background noise. We first employ learnable action queries to selectively emphasize action-relevant regions without requiring region-specific labels. Next, these queries are learned by a local action streaming branch that enables progressive query propagation. Moreover, we introduce a parameter-free feature interaction mechanism for effective multi-scale interaction between global and local features with minimal additional overhead. Extensive experiments demonstrate that FocusVideo achieves state-of-the-art performance across multiple action recognition datasets, validating its effectiveness and robustness in handling action-relevant details.

Diffusion models have demonstrated significant potential for generating high-quality images, audio, and videos. However, their iterative inference process entails substantial computational costs, limiting practical applications. Recently, researchers have introduced accelerated sampling methods that enable diffusion models to generate samples with far fewer timesteps than those used during training. Nonetheless, as the number of sampling steps decreases, the prediction errors significantly degrade the quality of generated outputs. Additionally, the exposure bias in diffusion models further amplifies these errors. To address these challenges, we leverage a manifold hypothesis to explore the exposure bias problem in depth. Based on this geometric perspective, we propose a manifold constraint that effectively reduces exposure bias during accelerated sampling of diffusion models. Notably, our method involves no additional training and requires only minimal hyperparameter tuning. Extensive experiments demonstrate the effectiveness of our approach, achieving a FID score of 15.60 with 10-step SDXL on MS-COCO, surpassing the baseline by a reduction of 2.57 in FID.

3D visual grounding allows an embodied agent to understand visual information in real-world 3D environments based on human instructions, which is crucial for embodied intelligence. Existing 3D visual grounding methods typically rely on separate encoders for different modalities (e.g., RGB images, text, and 3D point clouds), resulting in large and complex models that are inefficient to train. While some approaches use pre-trained 2D multi-modal models like CLIP for 3D tasks, they still struggle with aligning point cloud data to 2D encoders. As a result, these methods continue to depend on 3D encoders for feature extraction, further increasing model complexity and training inefficiency. In this paper, we propose a unified 2D pre-trained multi-modal network to process all three modalities (RGB images, text, and point clouds), significantly simplifying the architecture. By leveraging a 2D CLIP bi-modal model with adapter-based fine-tuning, this framework effectively adapts to the tri-modal setting, improving both adaptability and performance across modalities. Our Geometric-Aware 2D-3D Feature Recovery and Fusion (GARF) module is designed to fuse geometric multi-scale features from point clouds and images. We then integrate textual features for final modality fusion and introduce a multi-modal decoder to facilitate deep cross-modal understanding. Together, our method achieves unified feature extraction and fusion across the three modalities, enabling an end-to-end 3D visual grounding model. Compared to the baseline, our method reduces the number of trainable parameters by approximately 58%, while achieving a 6.52% improvement in the 3D detection task and a 6.25% improvement in the 3D visual grounding task.

With the rapid advancement of video generation models such as Sora, video quality assessment (VQA) is becoming increasingly crucial for selecting high-quality videos from large-scale datasets used in pre-training. Traditional VQA methods, typically producing single numerical scores, often lack comprehensiveness and interpretability. To address these challenges, we introduce MVQA-68K, a novel multi-dimensional VQA dataset comprising over 68,000 carefully annotated videos, covering seven essential quality dimensions: overall aesthetics, camera movement, dynamic degree, texture detail, composition, visual quality, and factual consistency. Each annotation includes detailed chain-of-thought reasoning to facilitate interpretability and comprehensive understanding. Extensive experiments demonstrate that MVQA-68K significantly enhances the performance of various multimodal large language models (MLLMs) on the VQA task, achieving state-of-the-art results not only on our internal test set (Fig.1) but also on public benchmarks including LSVQ-test, LSVQ-1080p, and LIVE-VQC. Meantime, incorporating explicit reasoning process during VQA training substantially boosts the zero-shot generalization. Code and dataset will be available at github: https://github.com/Controller01-ai/MVQA-68K.

Large language models (LLMs) have made significant advancements in natural language understanding. However, through that enormous semantic representation that the LLM has learnt, is it somehow possible for it to understand images as well? This work investigates this question. To enable the LLM to process images, we convert them into a representation given by Scalable Vector Graphics (SVG). To study what the LLM can do with this XML-based textual description of images, we test the LLM on three broad computer vision tasks: (i) visual reasoning and question answering, (ii) image classification under distribution shift, few-shot learning, and (iii) generating new images using visual prompting. Even though we do not naturally associate LLMs with any visual understanding capabilities, our results indicate that the LLM can often do a decent job in many of these tasks, potentially opening new avenues for research into LLMs' ability to understand image data. Our code, data, and models can be found here https://github.com/mu-cai/svg-11m.

Domain generalization studies the problem of training a model with samples from several domains (or distributions) and then testing the model with samples from a new, unseen domain. In this paper, we propose a novel approach for domain generalization that leverages recent advances in large vision-language models, specifically a CLIP teacher model, to train a smaller model that generalizes to unseen domains. The key technical contribution is a new type of regularization that requires the student’s learned image representations to be close to the teacher’s learned text representations obtained from encoding the corresponding text descriptions of images. We introduce two designs of the loss function, absolute and relative distance, which provide specific guidance on how the training process of the student model should be regularized. We evaluate our proposed method, dubbed RISE (Regularized Invariance with Semantic Embeddings), on various benchmark datasets, and show that it outperforms several state-of-the-art domain generalization methods. To our knowledge, our work is the first to leverage knowledge distillation using a large vision-language model for domain generalization. By incorporating text-based information, RISE improves the generalization capability of machine learning models.

Training with an emphasis on “hard-to-learn” components of the data has been proven as an effective method to improve the generalization of machine learning models, especially in the settings where robustness (e.g., generalization across distributions) is valued. Existing literature discussing this “hard-to-learn” concept are mainly expanded either along the dimension of the samples or the dimension of the features. In this paper, we aim to introduce a simple view merging these two dimensions, leading to a new, simple yet effective, heuristic to train machine learning models by emphasizing the worst-cases on both the sample and the feature dimensions. We name our method W2D following the concept of “Worst-case along Two Dimensions”. We validate the idea and demonstrate its empirical strength over standard benchmarks.

Convolution and self-attention are two powerful techniques for representation learning, and they are usually considered as two peer approaches that are distinct from each other. In this paper, we show that there exists a strong underlying relation between them, in the sense that the bulk of computations of these two paradigms are in fact done with the same operation. Specifically, we first show that a traditional convolution with kernel size k × k can be decomposed into k2individual 1 × 1 convolutions, followed by shift and summation operations. Then, we interpret the projections of queries, keys, and values in self-attention module as multiple 1 × 1 convolutions, followed by the computation of attention weights and aggregation of the values. Therefore, the first stage of both two modules comprises the similar operation. More importantly, the first stage contributes a dominant computation complexity (square of the channel size) comparing to the second stage. This observation naturally leads to an elegant integration of these two seemingly distinct paradigms, i.e., a mixed model that enjoys the benefit of both self-Attention and Convolution (ACmix), while having minimum compu-tational overhead compared to the pure convolution or self-attention counterpart. Extensive experiments show that our model achieves consistently improved results over com-petitive baselines on image recognition and downstream tasks. Code and pre-trained models will be released at https://github.com/LeapLabTHU/ACmix and https://gitee.com/mindspore/models.

Data augmentation has been proven to be an effective technique for developing machine learning models that are robust to known classes of distributional shifts (e.g., rotations of images), and alignment regularization is a technique often used together with data augmentation to further help the model learn representations invariant to the shifts used to augment the data. In this paper, motivated by a proliferation of options of alignment regularizations, we seek to evaluate the performances of several popular design choices along the dimensions of robustness and invariance, for which we introduce a new test procedure. Our synthetic experiment results speak to the benefits of squared ℓ2 norm regularization. Further, we also formally analyze the behavior of alignment regularization to complement our empirical study under assumptions we consider realistic. Finally, we test this simple technique we identify (worst-case data augmentation with squared ℓ2 norm alignment regularization) and show that the benefits of this method outrun those of the specially designed methods. We also release a software package in both TensorFlow and PyTorch for users to use the method with a couple of lines at https://github.com/jyanln/AlignReg.

Machine learning has demonstrated remarkable prediction accuracy over i.i.d data, but the accuracy often drops when tested with data from another distribution. In this paper, we aim to offer another view of this problem in a perspective assuming the reason behind this accuracy drop is the reliance of models on the features that are not aligned well with how a data annotator considers similar across these two datasets. We refer to these features as misaligned features. We extend the conventional generalization error bound to a new one for this setup with the knowledge of how the misaligned features are associated with the label. Our analysis offers a set of techniques for this problem, and these techniques are naturally linked to many previous methods in robust machine learning literature. We also compared the empirical strength of these methods demonstrated the performance when these previous techniques are combined, with implementation available.

Vision Transformers (ViT) have achieved remarkable success in large-scale image recognition. They split every 2D image into a fixed number of patches, each of which is treated as a token. Generally, representing an image with more tokens would lead to higher prediction accuracy, while it also results in drastically increased computational cost. To achieve a decent trade-off between accuracy and speed, the number of tokens is empirically set to 16x16 or 14x14. In this paper, we argue that every image has its own characteristics, and ideally the token number should be conditioned on each individual input. In fact, we have observed that there exist a considerable number of “easy” images which can be accurately predicted with a mere number of 4x4 tokens, while only a small fraction of “hard” ones need a finer representation. Inspired by this phenomenon, we propose a Dynamic Transformer to automatically configure a proper number of tokens for each input image. This is achieved by cascading multiple Transformers with increasing numbers of tokens, which are sequentially activated in an adaptive fashion at test time, i.e., the inference is terminated once a sufficiently confident prediction is produced. We further design efficient feature reuse and relationship reuse mechanisms across different components of the Dynamic Transformer to reduce redundant computations. Extensive empirical results on ImageNet, CIFAR-10, and CIFAR-100 demonstrate that our method significantly outperforms the competitive baselines in terms of both theoretical computational efficiency and practical inference speed. Code and pre-trained models (based on PyTorch and MindSpore) are available at https://github.com/blackfeather-wang/Dynamic-Vision-Transformer and https://github.com/blackfeather-wang/Dynamic-Vision-Transformer-MindSpore.

Classification and localization are two pillars of visual object detectors. However, in CNN-based detectors, these two modules are usually optimized under a fixed set of candidate (or anchor) bounding boxes. This configuration significantly limits the possibility to jointly optimize classification and localization. In this paper, we propose a Multiple Instance Learning (MIL) approach that selects anchors and jointly optimizes the two modules of a CNN-based object detector. Our approach, referred to as Multiple Anchor Learning (MAL), constructs anchor bags and selects the most representative anchors from each bag. Such an iterative selection process is potentially NP-hard to optimize. To address this issue, we solve MAL by repetitively depressing the confidence of selected anchors by perturbing their corresponding features. In an adversarial selection-depression manner, MAL not only pursues optimal solutions but also fully leverages multiple anchors/features to learn a detection model. Experiments show that MAL improves the baseline RetinaNet with significant margins on the commonly used MS-COCO object detection benchmark and achieves new state-of-the-art detection performance compared with recent methods.

We investigate the relationship between the frequency spectrum of image data and the generalization behavior of convolutional neural networks (CNN). We first notice CNN's ability in capturing the high-frequency components of images. These high-frequency components are almost imperceptible to a human. Thus the observation leads to multiple hypotheses that are related to the generalization behaviors of CNN, including a potential explanation for adversarial examples, a discussion of CNN's trade-off between robustness and accuracy, and some evidence in understanding training heuristics.

Weakly Supervised Object Detection (WSOD) has emerged as an effective tool to train object detectors using only the image-level category labels. However, without object-level labels, WSOD detectors are prone to detect bounding boxes on salient objects, clustered objects and discriminative object parts. Moreover, the image-level category labels do not enforce consistent object detection across different transformations of the same images. To address the above issues, we propose a Comprehensive Attention Self-Distillation (CASD) training approach for WSOD. To balance feature learning among all object instances, CASD computes the comprehensive attention aggregated from multiple transformations and feature layers of the same images. To enforce consistent spatial supervision on objects, CASD conducts self-distillation on the WSOD networks, such that the comprehensive attention is approximated simultaneously by multiple transformations and feature layers of the same images. CASD produces new state-of-the-art WSOD results on standard benchmarks such as PASCAL VOC 2007/2012 and MS-COCO.

Improving object detectors against occlusion, blur and noise is a critical step to deploy detectors in real applications. Since it is not possible to exhaust all image defects and occlusions through data collection, many researchers seek to generate occluded samples. The generated hard samples are either images or feature maps with coarse patches dropped out in the spatial dimensions. Significant overheads are required in generating hard samples and/or estimating drop-out patches using extra network branches. In this paper, we improve object detectors using a highly efficient and fine-grain mechanism called Inverted Attention (IA). Different from the original detector network that only focuses on the dominant part of objects, the detector network with IA iteratively inverts attention on feature maps which pushes the detector to discover new discriminative clues and puts more attention on complementary object parts, feature channels and even context. Our approach (1) operates along both the spatial and channels dimensions of the feature maps; (2) requires no extra training on hard samples, no extra network parameters for attention estimation, and no testing overheads. Experiments show that our approach consistently improved state-of-the-art detectors on benchmark databases.

Person re-identification (Re-ID) has undergone a rapid development with the blooming of deep neural network. Most methods are very easily affected by target misalignment and background clutter in the training process. In this paper, we propose a simple yet effective feedforward attention network to address the two mentioned problems, in which a novel consistent attention regularizer and an improved triplet loss are designed to learn foreground attentive features for person Re-ID. Specifically, the consistent attention regularizer aims to keep the deduced foreground masks similar from the low-level, mid-level and high-level feature maps. As a result, the network will focus on the foreground regions at the lower layers, which is benefit to learn discriminative features from the foreground regions at the higher layers. Last but not least, the improved triplet loss is introduced to enhance the feature learning capability, which can jointly minimize the intra-class distance and maximize the inter-class distance in each triplet unit. Experimental results on the Market1501, DukeMTMC-reID and CUHK03 datasets have shown that our method outperforms most of the state-of-the-art approaches.

Structured output support vector machine (SVM) based tracking algorithms have shown favorable performance recently. Nonetheless, the time-consuming candidate sampling and complex optimization limit their real-time applications. In this paper, we propose a novel large margin object tracking method which absorbs the strong discriminative ability from structured output SVM and speeds up by the correlation filter algorithm significantly. Secondly, a multimodal target detection technique is proposed to improve the target localization precision and prevent model drift introduced by similar objects or background noise. Thirdly, we exploit the feedback from high-confidence tracking results to avoid the model corruption problem. We implement two versions of the proposed tracker with the representations from both conventional hand-crafted and deep convolution neural networks (CNNs) based features to validate the strong compatibility of the algorithm. The experimental results demonstrate that the proposed tracker performs superiorly against several state-of-the-art algorithms on the challenging benchmark sequences while runs at speed in excess of 80 frames per second.

Nonlinear degradation trajectories are encountered frequently, and not all of them evolve homogeneously in practical systems. To take nonlinearity, heterogeneity, and the entire historical degradation data into account, we propose a nonlinear heterogeneous Wiener process model with an adaptive drift to characterize degradation trajectories. A state-space based method is employed to delineate our model. Due to the introduction of the adaptive drift, it is difficult to directly apply Kalman filter methods to update the distribution of the estimated degradation drift. To address this issue, we develop an online filtering algorithm based on Bayes' theorem. The expectation-maximization (EM) algorithm, as well as a novel Bayes'-theorem-based smoother, are adopted to estimate the unknown parameters in our model. Moreover, the distribution of the predicted remaining useful life (RUL) incorporating the complete distribution of the estimated degradation drift is achieved analytically. Finally, a simulation, and a case study are provided to validate the proposed approach.