ICCV 2023 Papers

Existing vector font generation approaches either struggle to preserve high-frequency corner details of the glyph or produce vector shapes that have redundant segments, which hinders their applications in practical scenarios. In this paper, we propose to learn vector fonts from pixelated font images utilizing a joint neural representation that consists of a signed distance field (SDF) and a probabilistic corner field (CF) to capture shape corner details. To achieve smooth shape interpolation on the learned shape manifold, we establish connections between the two fields for better alignment. We further design a vectorization process to extract high-quality and compact vector fonts from our joint neural representation. Experiments demonstrate that our method can generate more visually appealing vector fonts with a higher level of compactness compared to existing alternatives.

The quad-Bayer patterned image sensor has made significant improvements in spatial resolution over recent years due to advancements in image sensor technology. This has enabled single-shot high-dynamic-range (HDR) imaging using spatially varying multiple exposures. Popular methods for multi-exposure array sensors involve varying the gain of each exposure, but this does not effectively change the photoelectronic energy in each exposure. Consequently, HDR images produced using gain-based exposure variation may suffer from noise and details being saturated. To address this problem, we intend to use time-varying exposures in quad-Bayer patterned sensors. This approach allows long-exposure pixels to receive more photon energy than short- or middle-exposure pixels, resulting in higher-quality HDR images. However, time-varying exposures are not ideal for dynamic scenes and require an additional deghosting method. To tackle this issue, we propose a single-shot HDR demosaicing method that takes time-varying multiple exposures as input and jointly solves both the demosaicing and deghosting problems. Our method uses a feature-extraction module to handle mosaiced multiple exposures and a multiscale transformer module to register spatial displacements of multiple exposures and colors. We also created a dataset of quad-Bayer sensor input with time-varying exposures and trained our network using this dataset. Results demonstrate that our method outperforms baseline HDR reconstruction methods with both synthetic and real datasets. With our method, we can achieve high-quality HDR images in challenging lighting conditions.

In this study, we focus on the problem of 3D human mesh recovery from a single image under obscured conditions. Most state-of-the-art methods aim to improve 2D alignment technologies, such as spatial averaging and 2D joint sampling. However, they tend to neglect the crucial aspect of 3D alignment by improving 3D representations. Furthermore, recent methods struggle to separate target human from occlusion or background in crowded scenes as they optimize the 3D space of target human with 3D joint coordinates as local supervision. To address these issues, a desirable method would involve a framework for fusing 2D and 3D features and a strategy for optimizing the 3D space globally. Therefore, this paper presents 3D JOint contrastive learning with TRansformers (JOTR) framework for handling occluded 3D human mesh recovery. Our method includes an encoder-decoder transformer architecture to fuse 2D and 3D representations for achieving 2D&3D aligned results in a coarse-to-fine manner and a novel 3D joint contrastive learning approach for adding explicitly global supervision for the 3D feature space. The contrastive learning approach includes two contrastive losses: joint-to-joint contrast for enhancing the similarity of semantically similar voxels (i.e., human joints), and joint-to-non-joint contrast for ensuring discrimination from others (e.g., occlusions and background). Qualitative and quantitative analyses demonstrate that our method outperforms state-of-the-art competitors on both occlusion-specific and standard benchmarks, significantly improving the reconstruction of occluded humans.

Humans are good at recomposing novel objects, i.e they can identify commonalities between unknown objects from general structure to finer detail, an ability difficult to replicate by machines. We propose a framework, ISCO, to recompose an object using 3D superquadrics as semantic parts directly from 2D views without training a model that uses 3D supervision. To achieve this, we optimize the superquadric parameters that compose a specific instance of the object, comparing its rendered 3D view and 2D image silhouette. Our ISCO framework iteratively adds new superquadrics wherever the reconstruction error is high, abstracting first coarse regions and then finer details of the target object. With this simple coarse-to-fine inductive bias, ISCO provides consistent superquadrics for related object parts, despite not having any semantic supervision. Since ISCO does not train any neural network, it is also inherently robust to out of distribution objects. Experiments show that, compared to recent single instance superquadrics reconstruction approaches, ISCO provides consistently more accurate 3D reconstructions, even from images in the wild. Code available at https://github.com/ExplainableML/ISCO.

Image super-resolution (SR) has witnessed extensive neural network designs from CNN to transformer architectures. However, prevailing SR models suffer from prohibitive memory footprint and intensive computations, which limits further deployment on edge devices. This work investigates the potential of network pruning for super-resolution to take advantage of off-the-shelf network designs and reduce the underlying computational overhead. Two main challenges remain in applying pruning methods for SR. First, the widely-used filter pruning technique reflects limited granularity and restricted adaptability to diverse network structures. Second, existing pruning methods generally operate upon a pre-trained network for the sparse structure determination, hard to get rid of dense model training in the traditional SR paradigm. To address these challenges, we adopt unstructured pruning with sparse models directly trained from scratch. Specifically, we propose a novel Iterative Soft Shrinkage-Percentage (ISS-P) method by optimizing the sparse structure of a randomly initialized network at each iteration and tweaking unimportant weights with a small amount proportional to the magnitude scale on-the-fly. We observe that the proposed ISS-P can dynamically learn sparse structures adapting to the optimization process and preserve the sparse model's trainability by yielding a more regularized gradient throughput. Experiments on benchmark datasets demonstrate the effectiveness of the proposed ISS-P over diverse network architectures. Code is available at https://github.com/Jiamian-Wang/Iterative-Soft-Shrinkage-SR

With the emergence of powerful deep learning tools, more and more effective deep denoisers have advanced the field of image denoising. However, the huge progress made by these learning-based methods severely relies on large-scale and high-quality noisy/clean training pairs, which limits the practicality in real-world scenarios. To overcome this, researchers have been exploring self-supervised approaches that can denoise without paired data. However, the unavailable noise prior and inefficient feature extraction take these methods away from high practicality and precision. In this paper, we propose a Denoise-Corrupt-Denoise pipeline (DCD-Net) for self-supervised image denoising. Specifically, we design an iterative training strategy, which iteratively optimizes the denoiser and noise estimator, and gradually approaches high denoising performances using only single noisy images without any noise prior. The proposed self-supervised image denoising framework provides very competitive results compared with state-of-the-art methods on widely used synthetic and real-world image denoising benchmarks.

Recent leading zero-shot video object segmentation (ZVOS) works devote to integrating appearance and motion information by elaborately designing feature fusion modules and identically applying them in multiple feature stages. Our preliminary experiments show that with the strong long-range dependency modeling capacity of Transformer, simply concatenating the two modality features and feeding them to vanilla Transformers for feature fusion can distinctly benefit the performance but at a cost of heavy computation. Through further empirical analysis, we find that attention dependencies learned in Transformer in different stages exhibit completely different properties: global query-independent dependency in the low-level stages and semantic-specific dependency in the high-level stages. Motivated by the observations, we propose two Transformer variants: i) Context-Sharing Transformer (CST) that learns the global-shared contextual information within image frames with a lightweight computation. ii) Semantic Gathering-Scattering Transformer (SGST) that models the semantic correlation separately for the foreground and background and reduces the computation cost with a soft token merging mechanism. We apply CST and SGST for low-level and high-level feature fusions, respectively, formulating a level-isomerous Transformer framework for ZVOS task. Compared with the baseline that uses vanilla Transformers for multi-stage fusion, ours significantly increase the speed by 13 times and achieves new state-of-the-art ZVOS performance. Code is available at https://github.com/DLUT-yyc/Isomer.

We introduce DualMind, a generalist agent designed to tackle various decision-making tasks that addresses challenges posed by current methods, such as overfitting behaviors and dependence on task-specific fine-tuning. DualMind uses a novel "Dual-phase" training strategy that emulates how humans learn to act in the world. The model first learns fundamental common knowledge through a self-supervised objective tailored for control tasks and then learns how to make decisions based on different contexts through imitating behaviors conditioned on given prompts. DualMind can handle tasks across domains, scenes, and embodiments using just a single set of model weights and can execute zero-shot prompting without requiring task-specific fine-tuning. We evaluate DualMind on MetaWorld and Habitat through extensive experiments and demonstrate its superior generalizability compared to previous techniques, outperforming other generalist agents by over 50% and 70% on Habitat and MetaWorld, respectively. On the 45 tasks in MetaWorld, DualMind achieves over 30 tasks at a 90% success rate. Our source code is available at https://github.com/yunyikristy/DualMind.

In this paper, we propose to regularize ill-posed inverse problems using a deep hierarchical Variational AutoEncoder (HVAE) as an image prior. The proposed method synthesizes the advantages of i) denoiser-based Plug & Play approaches and ii) generative model based approaches to inverse problems. First, we exploit VAE properties to design an efficient algorithm that benefits from convergence guarantees of Plug-and-Play (PnP) methods. Second, our approach is not restricted to specialized datasets and the proposed PnP-HVAE model is able to solve image restoration problems on natural images of any size. Our experiments show that the proposed PnP-HVAE method is competitive with both SOTA denoiser-based PnP approaches, and other SOTA restoration methods based on generative models. The code for this project is available at https://github.com/jprost76/PnP-HVAE.

Video relation grounding (VRG) is a significant and challenging problem in the domains of cross-modal learning and video understanding. In this study, we introduce a novel approach called inverse compositional learning (ICL) for weakly-supervised video relation grounding. Our approach represents relations at both the holistic and partial levels, formulating VRG as a joint optimization problem that encompasses reasoning at both levels. For holistic-level reasoning, we propose an inverse attention mechanism and a compositional encoder to generate compositional relevance features. Additionally, we introduce an inverse loss to evaluate and learn the relevance between visual features and relation features. At the partial-level reasoning, we introduce a grounding by classification scheme. By leveraging the learned holistic-level features and partial-level features, we train the entire model in an end-to-end manner. We conduct evaluations on two challenging datasets and demonstrate the substantial superiority of our proposed method over state-of-the-art methods. Extensive ablation studies confirm the effectiveness of our approach.

We tackle the data scarcity challenge in few-shot point cloud recognition of 3D objects by using a joint prediction from a conventional 3D model and a well-pretrained 2D model. Surprisingly, such an ensemble, though seems trivial, has hardly been shown effective in recent 2D-3D models. We find out the crux is the less effective training for the "joint hard samples", which have high confidence prediction on different wrong labels, implying that the 2D and 3D models do not collaborate well. To this end, our proposed invariant training strategy, called INVJOINT, does not only emphasize the training more on the hard samples, but also seeks the invariance between the conflicting 2D and 3D ambiguous predictions. INVJOINT can learn more collaborative 2D and 3D representations for better ensemble. Extensive experiments on 3D shape classification with widely-adopted ModelNet10/40, ScanObjectNN and Toys4K, and shape retrieval with ShapeNet-Core validate the superiority of our INVJOINT.

Fair face recognition is all about learning invariant feature that generalizes to unseen faces in any demographic group. Unfortunately, face datasets inevitably capture the imbalanced demographic attributes that are ubiquitous in real-world observations, and the model learns biased feature that generalizes poorly in the minority group. We point out that the bias arises due to the confounding demographic attributes, which mislead the model to capture the spurious demographic-specific feature. The confounding effect can only be removed by causal intervention, which requires the confounder annotations. However, such annotations can be prohibitively expensive due to the diversity of the demographic attributes. To tackle this, we propose to generate diverse data partitions iteratively in an unsupervised fashion. Each data partition acts as a self-annotated confounder, enabling our Invariant Feature Regularization (INV-REG) to deconfound. INV-REG is orthogonal to existing methods, and combining INV-REG with two strong baselines (Arcface and CIFP) leads to new state-of-the-art that improves face recognition on a variety of demographic groups. Code is available at https://github.com/milliema/InvReg.

Continual Learning aims to learn a single model on a sequence of tasks without having access to data from previous tasks. The biggest challenge in the domain still remains catastrophic forgetting: a loss in performance on seen classes of earlier tasks. Some existing methods rely on an expensive replay buffer to store a chunk of data from previous tasks. This, while promising, becomes expensive when the number of tasks becomes large or data can not be stored for privacy reasons. As an alternative, prompt-based methods have been proposed that store the task information in a learnable prompt pool. This prompt pool instructs a frozen image encoder on how to solve each task. While the model faces a disjoint set of classes in each task in this setting, we argue that these classes can be encoded to the same embedding space of a pre-trained language encoder. In this work, we propose Language Guidance for Prompt-based Continual Learning (LGCL) as a plug-in for prompt-based methods. LGCL is model agnostic and introduces language guidance at the task level in the prompt pool and at the class level on the output feature of the vision encoder. We show with extensive experimentation that LGCL consistently improves the performance of prompt-based continual learning methods to set a new state-of-the-art. LGCL achieves these performance improvements without needing any additional learnable parameters.

Existing inverse rendering combined with neural rendering methods can only perform editable novel view synthesis on object-specific scenes, while we present intrinsic neural radiance fields, dubbed IntrinsicNeRF, which introduce intrinsic decomposition into the NeRF-based neural rendering method and can extend its application to room-scale scenes. Since intrinsic decomposition is a fundamentally under-constrained inverse problem, we propose a novel distance-aware point sampling and adaptive reflectance iterative clustering optimization method, which enables IntrinsicNeRF with traditional intrinsic decomposition constraints to be trained in an unsupervised manner, resulting in multi-view consistent intrinsic decomposition results. To cope with the problem that different adjacent instances of similar reflectance in a scene are incorrectly clustered together, we further propose a hierarchical clustering method with coarse-to-fine optimization to obtain a fast hierarchical indexing representation. It supports compelling real-time augmented applications such as recoloring and illumination variation. Extensive experiments and editing samples on both object-specific/room-scale scenes and synthetic/real-word data demonstrate that we can obtain consistent intrinsic decomposition results and high-fidelity novel view synthesis even for challenging sequences.

We propose a novel framework for interactive class-agnostic object counting, where a human user can interactively provide feedback to improve the accuracy of a counter. Our framework consists of two main components: a user-friendly visualizer to gather feedback and an efficient mechanism to incorporate it. In each iteration, we produce a density map to show the current prediction result, and we segment it into non-overlapping regions with an easily verifiable number of objects. The user can provide feedback by selecting a region with obvious counting errors and specifying the range for the estimated number of objects within it. To improve the counting result, we develop a novel adaptation loss to force the visual counter to output the predicted count within the user-specified range. For effective and efficient adaptation, we propose a refinement module that can be used with any density-based visual counter, and only the parameters in the refinement module will be updated during adaptation. Our experiments on two challenging class-agnostic object counting benchmarks, FSCD-LVIS and FSC-147, show that our method can reduce the mean absolute error of multiple state-of-the-art visual counters by roughly 30% to 40% with minimal user input. Our project can be found at https://yifehuang97.github.io/ICACountProjectPage/.

This paper proposes joint attention estimation in a single image. Different from related work in which only the gaze-related attributes of people are independently employed, (I) their locations and actions are also employed as contextual cues for weighting their attributes, and (ii) interactions among all of these attributes are explicitly modeled in our method. For the interaction modeling, we propose a novel Transformer-based attention network to encode joint attention as low-dimensional features. We introduce a specialized MLP head with positional embedding to the Transformer so that it predicts pixelwise confidence of joint attention for generating the confidence heatmap. This pixelwise prediction improves the heatmap accuracy by avoiding the ill-posed problem in which the high-dimensional heatmap is predicted from the low-dimensional features. The estimated joint attention is further improved by being integrated with general image-based attention estimation. Our method outperforms SOTA methods quantitatively in comparative experiments. Code: https://anonymous.4open.science/r/anonymized_codes-ECA4.

Interactive image segmentation enables annotators to efficiently perform pixel-level annotation for segmentation tasks. However, the existing interactive segmentation pipeline suffers from inefficient computations of interactive models because of the following two issues. First, annotators' later click is based on models' feedback of annotators' former click. This serial interaction is unable to utilize model's parallelism capabilities. Second, in each interaction step, the model handles the invariant image along with the sparse variable clicks, resulting in a process that's highly repetitive and redundant. For efficient computations, we propose a method named InterFormer that follows a new pipeline to address these issues. InterFormer extracts and preprocesses the computationally time-consuming part i.e. image processing from the existing process. Specifically, InterFormer employs a large vision transformer (ViT) on high-performance devices to preprocess images in parallel, and then uses a lightweight module called interactive multi-head self attention (I-MSA) for interactive segmentation. Furthermore, the I-MSA module's deployment on low-power devices extends the practical application of interactive segmentation. The I-MSA module utilizes the preprocessed features to efficiently response to the annotator inputs in real-time. The experiments on several datasets demonstrate the effectiveness of InterFormer, which outperforms previous interactive segmentation models in terms of computational efficiency and segmentation quality, achieve real-time high-quality interactive segmentation on CPU-only devices. The code is available at https://github.com/YouHuang67/InterFormer.

This paper addresses a novel task of anticipating 3D human-object interactions (HOIs). Most existing research on HOI synthesis lacks comprehensive whole-body interactions with dynamic objects, e.g., often limited to manipulating small or static objects. Our task is significantly more challenging, as it requires modeling dynamic objects with various shapes, capturing whole-body motion, and ensuring physically valid interactions. To this end, we propose InterDiff, a framework comprising two key steps: (i) interaction diffusion, where we leverage a diffusion model to encode the distribution of future human-object interactions; (ii) interaction correction, where we introduce a physics-informed predictor to correct denoised HOIs in a diffusion step. Our key insight is to inject prior knowledge that the interactions under reference with respect to contact points follow a simple pattern and are easily predictable. Experiments on multiple human-object interaction datasets demonstrate the effectiveness of our method for this task, capable of producing realistic, vivid, and remarkably long-term 3D HOI predictions.

Unsupervised anomaly detection and localization in images is a challenging problem, leading previous methods to attempt an easier supervised one-class classification formalization. Assuming training images to be realizations of the underlying image distribution, it follows that nominal patches from these realizations will be well associated between and represented across realizations. From this, we propose Inter-Realization Channels (InReaCh), a fully unsupervised method of detecting and localizing anomalies. InReaCh extracts high-confidence nominal patches from training data by associating them between realizations into channels, only considering channels with high spans and low spread as nominal. We then create our nominal model from the patches of these channels to test new patches against. InReaCh extracts nominal patches from the MVTec AD dataset with 99.9% precision, then archives 0.968 AUROC in localization and 0.923 AUROC in detection with corrupted training data, competitive with current state-of-the-art supervised one-class classification methods. We test our model up to 40% of training data containing anomalies with negligibly affected performance. The shift to fully unsupervised training simplifies dataset creation and broadens possible applications.