Vision-language models (VLMs), such as CLIP and ALIGN, are generally trained on datasets consisting of image-caption pairs obtained from the web. However, real-world multimodal datasets, such as healthcare data, are significantly more complex: each image (e.g. X-ray) is often paired with text (e.g. physician report) that describes many distinct attributes occurring in fine-grained regions of the image. We refer to these samples as exhibiting high pairwise complexity, since each image-text pair can be decomposed into a large number of region-attribute pairings. The extent to which VLMs can capture fine-grained relationships between image regions and textual attributes when trained on such data has not been previously evaluated. The first key contribution of this work is to demonstrate through systematic evaluations that as the pairwise complexity of the training dataset increases, standard VLMs struggle to learn region-attribute relationships, exhibiting performance degradations of up to 37% on retrieval tasks. In order to address this issue, we introduce ViLLA as our second key contribution. ViLLA, which is trained to capture fine-grained region-attribute relationships from complex datasets, involves two components: (a) a lightweight, self-supervised mapping model to decompose image-text samples into region-attribute pairs, and (b) a contrastive VLM to learn representations from generated region-attribute pairs. We demonstrate with experiments across four domains (synthetic, product, medical, and natural images) that ViLLA outperforms comparable VLMs on fine-grained reasoning tasks, such as zero-shot object detection (up to 3.6 AP50 points on COCO and 0.6 mAP points on LVIS) and retrieval (up to 14.2 R-Precision points).
论文检索
输入标题、作者或关键词，从 2,156 篇学术成果中精准定位
Shortcut-V2V: Compression Framework for Video-to-Video Translation Based on Temporal Redundancy Reduction
PDF ↗Video-to-video translation aims to generate video frames of a target domain from an input video. Despite its usefulness, the existing networks require enormous computations, necessitating their model compression for wide use. While there exist compression methods that improve computational efficiency in various image/video tasks, a generally applicable compression method for video-to-video translation has not been studied much. In response, we present Shortcut-V2V, a general-purpose compression framework for video-to-video translation. Shortcut-V2V avoids full inference for every neighboring video frame by approximating the intermediate features of a current frame from those of the previous frame. Moreover, in our framework, a newly-proposed block called AdaBD adaptively blends and deforms features of neighboring frames, which makes more accurate predictions of the intermediate features possible. We conduct quantitative and qualitative evaluations using well-known video-to-video translation models on various tasks to demonstrate the general applicability of our framework. The results show that Shortcut-V2V achieves comparable performance compared to the original video-to-video translation model while saving 3.2-5.7x computational cost and 7.8-44x memory at test time. Our code and videos are available at https://shortcut-v2v.github.io/.
CHORUS : Learning Canonicalized 3D Human-Object Spatial Relations from Unbounded Synthesized Images
PDF ↗We present a method for teaching machines to understand and model the underlying spatial common sense of diverse human-object interactions in 3D in a self-supervised way. This is a challenging task, as there exist specific manifolds of the interactions that can be considered human-like and natural, but the human pose and the geometry of objects can vary even for similar interactions. Such diversity makes the annotating task of 3D interactions difficult and hard to scale, which limits the potential to reason about that in a supervised way. One way of learning the 3D spatial relationship between humans and objects during interaction is by showing multiple 2D images captured from different viewpoints when humans interact with the same type of objects. The core idea of our method is to leverage a generative model that produces high-quality 2D images from an arbitrary text prompt input as an "unbounded" data generator with effective controllability and view diversity. Despite its imperfection of the image quality over real images, we demonstrate that the synthesized images are sufficient to learn the 3D human-object spatial relations. We present multiple strategies to leverage the synthesized images, including (1) the first method to leverage a generative image model for 3D human-object spatial relation learning; (2) a framework to reason about the 3D spatial relations from inconsistent 2D cues in a self-supervised manner via 3D occupancy reasoning with pose canonicalization; (3) semantic clustering to disambiguate different types of interactions with the same object types; and (4) a novel metric to assess the quality of 3D spatial learning of interaction. Project Page: https://jellyheadandrew.github.io/projects/chorus
TransTIC: Transferring Transformer-based Image Compression from Human Perception to Machine Perception
PDF ↗This work aims for transferring a Transformer-based image compression codec from human perception to machine perception without fine-tuning the codec. We propose a transferable Transformer-based image compression framework, termed TransTIC. Inspired by visual prompt tuning, TransTIC adopts an instance-specific prompt generator to inject instance-specific prompts to the encoder and task-specific prompts to the decoder. Extensive experiments show that our proposed method is capable of transferring the base codec to various machine tasks and outperforms the competing methods significantly. To our best knowledge, this work is the first attempt to utilize prompting on the low-level image compression task.
We introduce LDL, a fast and robust algorithm that localizes a panorama to a 3D map using line segments. LDL focuses on the sparse structural information of lines in the scene, which is robust to illumination changes and can potentially enable efficient computation. While previous line-based localization approaches tend to sacrifice accuracy or computation time, our method effectively observes the holistic distribution of lines within panoramic images and 3D maps. Specifically, LDL matches the distribution of lines with 2D and 3D line distance functions, which are further decomposed along principal directions of lines to increase the expressiveness. The distance functions provide coarse pose estimates by comparing the distributional information, where the poses are further optimized using conventional local feature matching. As our pipeline solely leverages line geometry and local features, it does not require costly additional training of line-specific features or correspondence matching. Nevertheless, our method demonstrates robust performance on challenging scenarios including object layout changes, illumination shifts, and large-scale scenes, while exhibiting fast pose search terminating within a matter of milliseconds. We thus expect our method to serve as a practical solution for line-based localization, and complement the well-established point-based paradigm.
Being able to perceive the semantics and the spatial structure of the environment is essential for visual navigation of a household robot. However, most existing works only employ visual backbones pre-trained either with independent images for classification or with self-supervised learning methods to adapt to the indoor navigation domain, both neglecting the spatial relationships that are essential to the learning of navigation. Inspired by the behavior that human naturally build semantically and spatially meaningful cognitive maps in their brain during navigation, in this paper, we propose a novel navigational-specific visual representation learning method by contrasting the agent's egocentric views and semantic maps (Ego^2-Map). We apply the visual transformer as the backbone encoder and train the model with data collected from the large-scale Habitat-Matterport3D environments. Ego^2-Map learning transfers the compact and rich information from a map, such as objects, structure and transition, to the agent's egocentric representations for navigation. Experiments show that agents using our learned representations on object-goal navigation outperforms recent visual pre-training methods. Moreover, our representations lead to a significant improvement in vision-and-language navigation in continuous environments for both high-level and low-level action spaces, achieving new state-of-the-art results of 47% SR and 41% SPL on the test server.
We introduce AiT, a unified output representation for various vision tasks, which is a crucial step towards general-purpose vision task solvers. Despite the challenges posed by the high-dimensional and task-specific outputs, we showcase the potential of using discrete representation (VQ-VAE) to model the dense outputs of many computer vision tasks as a sequence of discrete tokens. This is inspired by the established ability of VQ-VAE to conserve the structures spanning multiple pixels using few discrete codes. To that end, we present a modified shallower architecture for VQ-VAE that improves efficiency while keeping prediction accuracy. Our approach also incorporates uncertainty into the decoding process by using a soft fusion of the codebook entries, providing a more stable training process, which notably improved prediction accuracy. Our evaluation of AiT on depth estimation and instance segmentation tasks, with both continuous and discrete labels, demonstrates its superiority compared to other unified models. The code and models are available at https://github.com/SwinTransformer/AiT.
A triangular mesh is one of the most popular 3D data representations. As such, the deployment of deep neural networks for mesh processing is widely spread and is increasingly attracting more attention. However, neural networks are prone to adversarial attacks, where carefully crafted inputs impair the model's functionality. The need to explore these vulnerabilities is a fundamental factor in the future development of 3D-based applications. Recently, mesh attacks were studied on the semantic level, where classifiers are misled to produce wrong predictions. Nevertheless, mesh surfaces possess complex geometric attributes beyond their semantic meaning, and their analysis often includes the need to encode and reconstruct the geometry of the shape. We propose a novel framework for a geometric adversarial attack on a 3D mesh autoencoder. In this setting, an adversarial input mesh deceives the autoencoder by forcing it to reconstruct a different geometric shape at its output. The malicious input is produced by perturbing a clean shape in the spectral domain. Our method leverages the spectral decomposition of the mesh along with additional mesh-related properties to obtain visually credible results that consider the delicacy of surface distortions.
Autonomous driving requires accurate local scene understanding information. To this end, autonomous agents deploy object detection and online BEV lane graph extraction methods as a part of their perception stack. In this work, we propose an architecture and loss formulation to improve the accuracy of local lane graph estimates by using 3D object detection outputs. The proposed method learns to assign the objects to centerlines by considering the centerlines as cluster centers and the objects as data points to be assigned a probability distribution over the cluster centers. This training scheme ensures direct supervision on the relationship between lanes and objects, thus leading to better performance. The proposed method improves lane graph estimation substantially over state-of-the-art methods. The extensive ablations show that our method can achieve significant performance improvements by using the outputs of existing 3D object detection methods. Since our method uses the detection outputs rather than detection method intermediate representations, a single model of our method can use any detection method at test time. The code will be made publicly available.
Both indoor and outdoor environments are inherently structured and repetitive. Traditional modeling pipelines keep an asset library storing unique object templates, which is both versatile and memory efficient in practice. Inspired by this observation, we propose AssetField, a novel neural scene representation that learns a set of object-aware ground feature planes to represent the scene, where an asset library storing template feature patches can be constructed in an unsupervised manner. Unlike existing methods which require object masks to query spatial points for object editing, our ground feature plane representation offers a natural visualization of the scene in the bird-eye view, allowing a variety of operations (e.g. translation, duplication, deformation) on objects to configure a new scene. With the template feature patches, group editing is enabled for scenes with many recurring items to avoid repetitive work on object individuals. We show that AssetField not only achieves competitive performance for novel-view synthesis but also generates realistic renderings for new scene configurations.
Generating full-body and multi-genre dance sequences from given music is a challenging task, due to the limitations of existing datasets and the inherent complexity of the fine-grained hand motion and dance genres. To address these problems, we propose FineDance, which contains 14.6 hours of music-dance paired data, with fine-grained hand motions, fine-grained genres (22 dance genres), and accurate posture. To the best of our knowledge, FineDance is the largest music-dance paired dataset with the most dance genres. Additionally, to address monotonous and unnatural hand movements existing in previous methods, we propose a full-body dance generation network, which utilizes the diverse generation capabilities of the diffusion model to solve monotonous problems, and use expert nets to solve unreal problems. To further enhance the genrematching and long-term stability of generated dances, we propose a Genre&Coherent aware Retrieval Module. Besides, we propose a new metric named Genre Matching Score to measure the genre matching between dance and music. Quantitative and qualitative experiments demonstrate the quality of FineDance, and the state-of-the-art performance of FineNet.
We propose a novel and efficient method for reconstructing manifold surfaces from point clouds. Unlike previous approaches that use dense implicit reconstructions or piecewise approximations and overlook inherent structures like quadrics in CAD models, our method faithfully preserves these quadric structures by assembling primitives. To achieve high-quality primitive extraction, we use a variational shape approximation, followed by a mesh arrangement for space partitioning and candidate primitive patches generation. We then introduce an effective pruning mechanism to classify candidate primitive patches as active or inactive, and further prune inactive patches to reduce the search space and speed up surface extraction significantly. Finally, the optimal active patches are computed by a binary linear programming and assembled as manifold and watertight surfaces. We perform extensive experiments on a wide range of CAD objects to validate its effectiveness.
Recently, GAN has successfully contributed to making single-image super-resolution (SISR) methods produce more realistic images. However, natural images have complex distribution in the real world, and a single classifier in the discriminator may not have enough capacity to classify real and fake samples, making the preceding SR network generate unpleasing noise and artifacts. To solve the problem, we propose a novel content-aware local GAN framework, CAL-GAN, which processes a large and complicated distribution of real-world images by dividing them into smaller subsets based on similar contents. Our mixture of classifiers (MoC) design allocates different super-resolved patches to corresponding expert classifiers. Additionally, we introduce novel routing and orthogonality loss terms so that different classifiers can handle various contents and learn separable features. By feeding similar distributions into the corresponding specialized classifiers, CAL-GAN enhances the representation power of existing super-resolution models, achieving state-of-the-art perceptual performance on standard benchmarks and real-world images without modifying the generator-side architecture.
Understanding the emotions in text and presenting them visually is a very challenging problem that requires a deep understanding of natural language and high-quality image synthesis simultaneously. In this work, we propose Affective Image Filter (AIF), a novel model that is able to understand the visually-abstract emotions from the text and reflect them to visually-concrete images with appropriate colors and textures. We build our model based on the multi-modal transformer architecture, which unifies both images and texts into tokens and encodes the emotional prior knowledge. Various loss functions are proposed to understand complex emotions and produce appropriate visualization. In addition, we collect and contribute a new dataset with abundant aesthetic images and emotional texts for training and evaluating the AIF model. We carefully design four quantitative metrics and conduct a user study to comprehensively evaluate the performance, which demonstrates our AIF model outperforms state-of-the-art methods and could evoke specific emotional responses from human observers.
3D scene reconstruction is a long-standing vision task. Existing approaches can be categorized into geometry-based and learning-based methods. The former leverages multi-view geometry but may face catastrophic failures due to the reliance on accurate pixel correspondence across views, while the latter mitigates these issues by learning 2D or 3D representation directly. However, without a large-scale video or 3D training data, it can hardly be generalized to diverse real-world scenarios due to the presence of tens of millions or even billions of optimization parameters in the deep network. Recently, robust monocular depth estimation models trained with large-scale datasets have been proven to possess weak 3D geometry prior, but they are insufficient for reconstruction due to the unknown camera parameters, the affine-invariant property, and inter-frame inconsistency. To address these issues, we propose a novel test-time optimization approach that can transfer the robustness of affine-invariant depth models such as LeReS to challenging diverse scenes while ensuring inter-frame consistency, with only dozens of parameters to optimize per video frame. Specifically, our approach involves freezing the pre-trained affine-invariant depth model's depth predictions, rectifying them by optimizing the unknown scale-shift values with a geometric consistency alignment module, and employing the resulting scale-consistent depth maps to robustly obtain camera poses and achieve dense scene reconstruction, even in low-texture regions. Experiments show that our method achieves state-of-the-art cross-dataset reconstruction on five zero-shot testing datasets. Code is available at: https://aim-uofa.github.io/FrozenRecon/
Deep Learning models have shown remarkable performance in a broad range of vision tasks. However, they are often vulnerable against domain shifts at test-time. Test-time training (TTT) methods have been developed in an attempt to mitigate these vulnerabilities, where a secondary task is solved at training time simultaneously with the main task, to be later used as an self-supervised proxy task at test-time. In this work, we propose a novel unsupervised TTT technique based on the maximization of Mutual Information between multi-scale feature maps and a discrete latent representation, which can be integrated to the standard training as an auxiliary clustering task. Experimental results demonstrate competitive classification performance on different popular test-time adaptation benchmarks. The code can be found at: https://github.com/dosowiechi/ClusT3.git
Domain adaptation (DA) aims to transfer knowledge from a label-rich source domain to a related but label-scarce target domain. Recently, increasing research has focused on exploring data structure of the target domain. In light of the recent success of Instance Discrimination Contrastive (IDCo) loss in self-supervised learning, we try directly applying it to domain adaptation tasks. However, the improvement is very limited, which motivates us to rethink its underlying limitations for domain adaptation tasks. An intuitive limitation is that a pair of samples belonging to the same class could be treated as negatives. Here we argue that using low-confidence samples to construct positive and negative pairs can alleviate this issue and is more suitable for IDCo loss. Another limitation is that IDCo loss cannot capture enough semantic information. We address this by introducing domain-invariant and accurate semantic information from classifier weights and input data. Specifically, we propose a class relationship enhanced features. It uses probability weighted class prototpyes as the input features of IDCo loss, which can implicitly transfer the domain-invariant class relationship. We further propose a target-dominated cross-domain mixup that can incorporate accurate semantic information from the source domain. We evaluate the proposed method in unsupervised DA and other DA settings, and extensive experimental results reveal that our method can make IDCo loss more effective and achieve state-of-the-art performance.
This paper presents an improved DETR detector that maintains a "plain" nature: using a single-scale feature map and global cross-attention calculations without specific locality constraints, in contrast to previous leading DETR-based detectors that reintroduce architectural inductive biases of multi-scale and locality into the decoder. We show that two simple technologies are surprisingly effective within a plain design to compensate for the lack of multi-scale feature maps and locality constraints. The first is a box-to-pixel relative position bias (BoxRPB) term added to the cross-attention formulation, which well guides each query to attend to the corresponding object region while also providing encoding flexibility. The second is masked image modeling (MIM)-based backbone pre-training which helps learn representation with fine-grained localization ability and proves crucial for remedying dependencies on the multi-scale feature maps. By incorporating these technologies and recent advancements in training and problem formation, the improved "plain" DETR showed exceptional improvements over the original DETR detector. By leveraging the Object365 dataset for pre-training, it achieved 63.9 mAP accuracy using a Swin-L backbone, which is highly competitive with state-of-the-art detectors which all heavily rely on multi-scale feature maps and region-based feature extraction. Code will be available at https://github.com/impiga/Plain-DETR.
Although Neural Radiance Fields (NeRF) is popular in the computer vision community recently, registering multiple NeRFs has yet to gain much attention. Unlike the existing work, NeRF2NeRF, which is based on traditional optimization methods and needs human annotated keypoints, we propose DReg-NeRF to solve the NeRF registration problem on object-centric scenes without human intervention. After training NeRF models, our DReg-NeRF first extracts features from the occupancy grid in NeRF. Subsequently, our DReg-NeRF utilizes a transformer architecture with self-attention and cross-attention layers to learn the relations between pairwise NeRF blocks. In contrast to state-of-the-art (SOTA) point cloud registration methods, the decoupled correspondences are supervised by surface fields without any ground truth overlapping labels. We construct a novel view synthesis dataset with 1,700+ 3D objects obtained from Objaverse to train our network. When evaluated on the test set, our proposed method beats the SOTA point cloud registration methods by a large margin with a mean RPE = 9.67* and a mean RTE = 0.038. Our code is available at https://github.com/AIBluefisher/DReg-NeRF.
Transparent objects are commonly seen in indoor scenes but are hard to estimate. Currently, commercial depth cameras face difficulties in estimating the depth of transparent objects due to the light reflection and refraction on their surface. As a result, they tend to make a noisy and incorrect depth value for transparent objects. These incorrect depth data make the traditional RGB-D SLAM method fails in reconstructing the scenes that contain transparent objects. An exact depth value of the transparent object is required to restore in advance and it is essential that the depth value of the transparent object must keep consistent in different views, or the reconstruction result will be distorted. Previous depth prediction methods of transparent objects can restore these missing depth values but none of them can provide a good result in reconstruction due to the inconsistency prediction. In this work, we propose a real-time reconstruction method using a novel stereo-based depth prediction network to keep the consistency of depth prediction in a sequence of images. Because there is no video dataset about transparent objects currently to train our model, we construct a synthetic RGB-D video dataset with different transparent objects. Moreover, to test generalization capability, we capture video from real scenes using the RealSense D435i RGB-D camera. We compare the metrics on our dataset and SLAM reconstruction results in both synthetic scenes and real scenes with the previous methods. Experiments show our significant improvement in accuracy on depth prediction and scene reconstruction.