Jiawang Bian

Shuai Chen, Yash Bhalgat, Xinghui Li et al. · bytedance, oxford

Absolute Pose Regression (APR) methods use deep neural networks to directly regress camera poses from RGB images. However, the predominant APR architectures only rely on 2D operations during inference, resulting in limited accuracy of pose estimation due to the lack of 3D geometry constraints or priors. In this work, we propose a test-time refinement pipeline that leverages implicit geometric constraints using a robust feature field to enhance the ability of APR methods to use 3D information during inference. We also introduce a novel Neural Feature Synthesizer (NeFeS) model, which encodes 3D geometric features during training and directly renders dense novel view features at test time to refine APR methods. To enhance the robustness of our model, we introduce a feature fusion module and a progressive training strategy. Our proposed method achieves state-of-the-art single-image APR accuracy on indoor and outdoor datasets.

Sensen Gao, Zhaoqing Wang, Qihang Cao et al.

Existing diffusion-based 3D scene generation methods primarily operate in 2D image/video latent spaces, which makes maintaining cross-view appearance and geometric consistency inherently challenging. To bridge this gap, we present OneWorld, a framework that performs diffusion directly within a coherent 3D representation space. Central to our approach is the 3D Unified Representation Autoencoder (3D-URAE); it leverages pretrained 3D foundation models and augments their geometry-centric nature by injecting appearance and distilling semantics into a unified 3D latent space. Furthermore, we introduce token-level Cross-View-Correspondence (CVC) consistency loss to explicitly enforce structural alignment across views, and propose Manifold-Drift Forcing (MDF) to mitigate train-inference exposure bias and shape a robust 3D manifold by mixing drifted and original representations. Comprehensive experiments demonstrate that OneWorld generates high-quality 3D scenes with superior cross-view consistency compared to state-of-the-art 2D-based methods. Our code will be available at https://github.com/SensenGao/OneWorld.

Haoyuan Li, Yanpeng Zhou, Yufei Gao et al.

Remarkable progress in 2D Vision-Language Models (VLMs) has spurred interest in extending them to 3D settings for tasks like 3D Question Answering, Dense Captioning, and Visual Grounding. Unlike 2D VLMs that typically process images through an image encoder, 3D scenes, with their intricate spatial structures, allow for diverse model architectures. Based on their encoder design, this paper categorizes recent 3D VLMs into 3D object-centric, 2D image-based, and 3D scene-centric approaches. Despite the architectural similarity of 3D scene-centric VLMs to their 2D counterparts, they have exhibited comparatively lower performance compared with the latest 3D object-centric and 2D image-based approaches. To understand this gap, we conduct an in-depth analysis, revealing that 3D scene-centric VLMs show limited reliance on the 3D scene encoder, and the pre-train stage appears less effective than in 2D VLMs. Furthermore, we observe that data scaling benefits are less pronounced on larger datasets. Our investigation suggests that while these models possess cross-modal alignment capabilities, they tend to over-rely on linguistic cues and overfit to frequent answer distributions, thereby diminishing the effective utilization of the 3D encoder. To address these limitations and encourage genuine 3D scene understanding, we introduce a novel 3D Relevance Discrimination QA dataset designed to disrupt shortcut learning and improve 3D understanding. Our findings highlight the need for advanced evaluation and improved strategies for better 3D understanding in 3D VLMs.

Huangying Zhan, Chamara Saroj Weerasekera, Jiawang Bian et al.

In this work we present a monocular visual odometry (VO) algorithm which leverages geometry-based methods and deep learning. Most existing VO/SLAM systems with superior performance are based on geometry and have to be carefully designed for different application scenarios. Moreover, most monocular systems suffer from scale-drift issue.Some recent deep learning works learn VO in an end-to-end manner but the performance of these deep systems is still not comparable to geometry-based methods. In this work, we revisit the basics of VO and explore the right way for integrating deep learning with epipolar geometry and Perspective-n-Point (PnP) method. Specifically, we train two convolutional neural networks (CNNs) for estimating single-view depths and two-view optical flows as intermediate outputs. With the deep predictions, we design a simple but robust frame-to-frame VO algorithm (DF-VO) which outperforms pure deep learning-based and geometry-based methods. More importantly, our system does not suffer from the scale-drift issue being aided by a scale consistent single-view depth CNN. Extensive experiments on KITTI dataset shows the robustness of our system and a detailed ablation study shows the effect of different factors in our system.

Yun Liu, Yu Qiu, Le Zhang et al.

Recent progress on salient object detection mainly aims at exploiting how to effectively integrate convolutional side-output features in convolutional neural networks (CNN). Based on this, most of the existing state-of-the-art saliency detectors design complex network structures to fuse the side-output features of the backbone feature extraction networks. However, should the fusion strategies be more and more complex for accurate salient object detection? In this paper, we observe that the contexts of a natural image can be well expressed by a high-to-low self-learning of side-output convolutional features. As we know, the contexts of an image usually refer to the global structures, and the top layers of CNN usually learn to convey global information. On the other hand, it is difficult for the intermediate side-output features to express contextual information. Here, we design an hourglass network with intermediate supervision to learn contextual features in a high-to-low manner. The learned hierarchical contexts are aggregated to generate the hybrid contextual expression for an input image. At last, the hybrid contextual features can be used for accurate saliency estimation. We extensively evaluate our method on six challenging saliency datasets, and our simple method achieves state-of-the-art performance under various evaluation metrics. Code will be released upon paper acceptance.

JiaWang Bian, Ruihan Yang, Yun Liu et al.

Feature matching is one of the most fundamental and active research areas in computer vision. A comprehensive evaluation of feature matchers is necessary, since it would advance both the development of this field and also high-level applications such as Structure-from-Motion or Visual SLAM. However, to the best of our knowledge, no previous work targets the evaluation of feature matchers while they only focus on evaluating feature detectors and descriptors. This leads to a critical absence in this field that there is no standard datasets and evaluation metrics to evaluate different feature matchers fairly. To this end, we present the first uniform feature matching benchmark to facilitate the evaluation of feature matchers. In the proposed benchmark, matchers are evaluated in different aspects, involving matching ability, correspondence sufficiency, and efficiency. Also, their performances are investigated in different scenes and in different matching types. Subsequently, we carry out an extensive evaluation of different state-of-the-art matchers on the benchmark and make in-depth analyses based on the reported results. This can be used to design practical matching systems in real applications and also advocates the potential future research directions in the field of feature matching.

Yun Liu, Yujun Shi, JiaWang Bian et al.

Deep learning stands at the forefront in many computer vision tasks. However, deep neural networks are usually data-hungry and require a huge amount of well-annotated training samples. Collecting sufficient annotated data is very expensive in many applications, especially for pixel-level prediction tasks such as semantic segmentation. To solve this fundamental issue, we consider a new challenging vision task, Internetly supervised semantic segmentation, which only uses Internet data with noisy image-level supervision of corresponding query keywords for segmentation model training. We address this task by proposing the following solution. A class-specific attention model unifying multiscale forward and backward convolutional features is proposed to provide initial segmentation "ground truth". The model trained with such noisy annotations is then improved by an online fine-tuning procedure. It achieves state-of-the-art performance under the weakly-supervised setting on PASCAL VOC2012 dataset. The proposed framework also paves a new way towards learning from the Internet without human interaction and could serve as a strong baseline therein. Code and data will be released upon the paper acceptance.

Yun Liu, Ming-Ming Cheng, Deng-Ping Fan et al.

Semantic edge detection (SED), which aims at jointly extracting edges as well as their category information, has far-reaching applications in domains such as semantic segmentation, object proposal generation, and object recognition. SED naturally requires achieving two distinct supervision targets: locating fine detailed edges and identifying high-level semantics. Our motivation comes from the hypothesis that such distinct targets prevent state-of-the-art SED methods from effectively using deep supervision to improve results. To this end, we propose a novel fully convolutional neural network using diverse deep supervision (DDS) within a multi-task framework where bottom layers aim at generating category-agnostic edges, while top layers are responsible for the detection of category-aware semantic edges. To overcome the hypothesized supervision challenge, a novel information converter unit is introduced, whose effectiveness has been extensively evaluated on SBD and Cityscapes datasets.

JiaWang Bian, Le Zhang, Yun Liu et al.

Image matching approaches have been widely used in computer vision applications in which the image-level matching performance of matchers is critical. However, it has not been well investigated by previous works which place more emphases on evaluating local features. To this end, we present a uniform benchmark with novel evaluation metrics and a large-scale dataset for evaluating the overall performance of image matching methods. The proposed metrics are application-oriented as they emphasize application requirements for matchers. The dataset contains two portions for benchmarking video frame matching and unordered image matching separately, where each portion consists of real-world image sequences and each sequence has a specific attribute. Subsequently, we carry out a comprehensive performance evaluation of different state-of-the-art methods and conduct in-depth analyses regarding various aspects such as application requirements, matching types, and data diversity. Moreover, we shed light on how to choose appropriate approaches for different applications based on empirical results and analyses. Conclusions in this benchmark can be used as general guidelines to design practical matching systems and also advocate potential future research directions in this field.