Li Zhang

Yurui Chen, Chun Gu, Feihu Zhang et al.

Neural Radiance Fields (NeRF) have been proposed for photorealistic novel view rendering. However, it requires many different views of one scene for training. Moreover, it has poor generalizations to new scenes and requires retraining or fine-tuning on each scene. In this paper, we develop a new NeRF model for novel view synthesis using only a single image as input. We propose to combine the (coarse) planar rendering and the (fine) volume rendering to achieve higher rendering quality and better generalizations. We also design a depth teacher net that predicts dense pseudo depth maps to supervise the joint rendering mechanism and boost the learning of consistent 3D geometry. We evaluate our method on three challenging datasets. It outperforms state-of-the-art single-view NeRFs by achieving 5$\sim$20\% improvements in PSNR and reducing 20$\sim$50\% of the errors in the depth rendering. It also shows excellent generalization abilities to unseen data without the need to fine-tune on each new scene.

Filippo Aleotti, Fabio Tosi, Li Zhang et al.

In many fields, self-supervised learning solutions are rapidly evolving and filling the gap with supervised approaches. This fact occurs for depth estimation based on either monocular or stereo, with the latter often providing a valid source of self-supervision for the former. In contrast, to soften typical stereo artefacts, we propose a novel self-supervised paradigm reversing the link between the two. Purposely, in order to train deep stereo networks, we distill knowledge through a monocular completion network. This architecture exploits single-image clues and few sparse points, sourced by traditional stereo algorithms, to estimate dense yet accurate disparity maps by means of a consensus mechanism over multiple estimations. We thoroughly evaluate with popular stereo datasets the impact of different supervisory signals showing how stereo networks trained with our paradigm outperform existing self-supervised frameworks. Finally, our proposal achieves notable generalization capabilities dealing with domain shift issues. Code available at https://github.com/FilippoAleotti/Reversing

Renyuan Peng, Xinyue Cai, Hang Xu et al.

Understanding road structures is crucial for autonomous driving. Intricate road structures are often depicted using lane graphs, which include centerline curves and connections forming a Directed Acyclic Graph (DAG). Accurate extraction of lane graphs relies on precisely estimating vertex and edge information within the DAG. Recent research highlights Transformer-based language models' impressive sequence prediction abilities, making them effective for learning graph representations when graph data are encoded as sequences. However, existing studies focus mainly on modeling vertices explicitly, leaving edge information simply embedded in the network. Consequently, these approaches fall short in the task of lane graph extraction. To address this, we introduce LaneGraph2Seq, a novel approach for lane graph extraction. It leverages a language model with vertex-edge encoding and connectivity enhancement. Our serialization strategy includes a vertex-centric depth-first traversal and a concise edge-based partition sequence. Additionally, we use classifier-free guidance combined with nucleus sampling to improve lane connectivity. We validate our method on prominent datasets, nuScenes and Argoverse 2, showcasing consistent and compelling results. Our LaneGraph2Seq approach demonstrates superior performance compared to state-of-the-art techniques in lane graph extraction.

Li Zhang, Quanhong Wang, Haihua Lu et al.

Deep neural networks have shown excellent performance in stereo matching task. Recently CNN-based methods have shown that stereo matching can be formulated as a supervised learning task. However, less attention is paid on the fusion of contextual semantic information and details. To tackle this problem, we propose a network for disparity estimation based on abundant contextual details and semantic information, called Multi-scale Features Network (MSFNet). First, we design a new structure to encode rich semantic information and fine-grained details by fusing multi-scale features. And we combine the advantages of element-wise addition and concatenation, which is conducive to merge semantic information with details. Second, a guidance mechanism is introduced to guide the network to automatically focus more on the unreliable regions. Third, we formulate the consistency check as an error map, obtained by the low stage features with fine-grained details. Finally, we adopt the consistency checking between the left feature and the synthetic left feature to refine the initial disparity. Experiments on Scene Flow and KITTI 2015 benchmark demonstrated that the proposed method can achieve the state-of-the-art performance.