Interpretable Vision Transformers in Monocular Depth Estimation via SVDA
Dept. Electrical and Computer Engineering（电子与计算机工程系）；Democritus University of Thrace（德米特里乌斯大学）；Athena Research Center（雅典研究中心）；University Campus at Kimmeria（基米里亚大学校园）
Vasileios Arampatzakis, George Pavlidis, Nikolaos Mitianoudis, Nikos Papamarkos
备注：
8 pages, 2 figures, submitted to CVPR Conference 2026
Monocular depth estimation is a central problem in computer vision with applications in robotics, AR, and autonomous driving, yet the self-attention mechanisms that drive modern Transformer architectures remain opaque. We introduce SVD-Inspired Attention (SVDA) into the Dense Prediction Transformer (DPT), providing the first spectrally structured formulation of attention for dense prediction tasks. SVDA decouples directional alignment from spectral modulation by embedding a learnable diagonal matrix into normalized query-key interactions, enabling attention maps that are intrinsically interpretable rather than post-hoc approximations. Experiments on KITTI and NYU-v2 show that SVDA preserves or slightly improves predictive accuracy while adding only minor computational overhead. More importantly, SVDA unlocks six spectral indicators that quantify entropy, rank, sparsity, alignment, selectivity, and robustness. These reveal consistent cross-dataset and depth-wise patterns in how attention organizes during training, insights that remain inaccessible in standard Transformers. By shifting the role of attention from opaque mechanism to quantifiable descriptor, SVDA redefines interpretability in monocular depth estimation and opens a principled avenue toward transparent dense prediction models.
Interpretable Vision Transformers in Image Classification via SVDA
Dept. Electrical and Computer Engineering（电子与计算机工程系）；Democritus University of Thrace（德摩克利特大学）；Athena Research Center（雅典研究中心）；University Campus at Kimmeria（基米里亚大学校园）
Sun Yat-sen University（中山大学）；NEC Laboratories America（美国 NEC 实验室）；National Chiao Tung University（国立交通大学）
Deep networks for Monocular Depth Estimation (MDE) have achieved promising performance recently and it is of great importance to further understand the interpretability of these networks. Existing methods attempt to provide posthoc explanations by investigating visual cues, which may not explore the internal representations learned by deep networks. In this paper, we find that some hidden units of the network are selective to certain ranges of depth, and thus such behavior can be served as a way to interpret the internal representations. Based on our observations, we quantify the interpretabil…
Beijing Jiaotong University（北京交通大学）；Institute of Information Science, Beijing Jiaotong University（北京交通大学信息科学研究所）
For a monocular 360 image, depth estimation is a challenging because the distortion increases along the latitude. To perceive the distortion, existing methods devote to designing a deep and complex network architecture. In this paper, we provide a new perspective that constructs an interpretable and sparse representation for a 360 image. Considering the importance of the geometric structure in depth estimation, we utilize the contourlet transform to capture an explicit geometric cue in the spectral domain and integrate it with an implicit cue in the spatial domain. Specifically, we propose a n…
Robust 3D Alignment of Generative Reconstructions via Partial Monocular Observations
School of Computer Science, Shanghai Jiao Tong University（上海交通大学计算机科学与技术学院）；School of Computer Science and Engineering, Nanjing University of Science and Technology（南京理工大学计算机科学与工程学院）；School of Engineering and Applied Science, University of Pennsylvania（宾夕法尼亚大学工程与应用科学学院）；School of Tech Electrical and Information Engineering, Tianjin University（天津大学电气与信息工程学院）；Independent Researcher（独立研究员）；School of Electronic Engineering and Information Science, University of Science and Technology of China（中国科学技术大学电子工程与信息科学学院）；Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）；School of Robotics, Wuhan University（武汉大学机器人学院）；Xiaomi EV（小米汽车）
University of Science and Technology of China（中国科学技术大学）；Image Derivative Inc（Image Derivative公司）
We present a novel semantic model for human head defined with neural radiance field. The 3D-consistent head model consist of a set of disentangled and interpretable bases, and can be driven by low-dimensional expression coefficients. Thanks to the powerful representation ability of neural radiance field, the constructed model can represent complex facial attributes including hair, wearings, which can not be represented by traditional mesh blendshape. To construct the personalized semantic facial model, we propose to define the bases as several multi-level voxel fields. With a short monocular R…
Peking University（北京大学）；Advanced Institute of Information Technology, Peking University（北京大学信息工程高等研究院）；Information Technology R&D Innovation Center of Peking University（北京大学信息技术研发创新中心）
We present two versatile methods to generally enhance self-supervised monocular depth estimation (MDE) models. The high generalizability of our methods is achieved by solving the fundamental and ubiquitous problems in photometric loss function. In particular, from the perspective of spatial frequency, we first propose Ambiguity-Masking to suppress the incorrect supervision under photometric loss at specific object boundaries, the cause of which could be traced to pixel-level ambiguity. Second, we present a novel frequency-adaptive Gaussian low-pass filter, designed to robustify the photometric…
University College London（伦敦大学学院）；Snap Inc.（Snap 公司）
Monocular 3D reconstruction of articulated object categories is challenging due to the lack of training data and the inherent ill-posedness of the problem. In this work we use video self-supervision, forcing the consistency of consecutive 3D reconstructions by a motion-based cycle loss. This largely improves both optimization-based and learning-based 3D mesh reconstruction. We further introduce an interpretable model of 3D template deformations that controls a 3D surface through the displacement of a small number of local, learnable handles. We formulate this operation as a structured layer re…
Sparse-View Interpretable 3D Animal Behavior Representations for Neural Encoding and Decoding
Harvard University（哈佛大学）；University of Cambridge（剑桥大学）；Northwestern University（西北大学）；Cold Spring Harbor Laboratory（冷泉港实验室）；Columbia University（哥伦比亚大学）；University of Science and Technology of China（中国科学技术大学）
The Midas Touch for Metric Depth: Interpretable Metricization for Generalizable Depth Perception
Shanghai Research Institute for Intelligent Autonomous Systems, Tongji University（同济大学上海智能自主系统研究院）；College of Electronic & Information Engineering, Tongji University（同济大学电子与信息工程学院）；The Chinese University of Hong Kong（香港中文大学）；Shanghai Institute of Intelligent Science and Technology（上海智能科学技术研究院）
Technical University of Munich（慕尼黑工业大学）；German Aerospace Center（德国航空航天中心）
Monocular height estimation (MHE) from remote sensing imagery has high potential in generating 3D city models efficiently for a quick response to natural disasters. Most existing works pursue higher performance. However, there is little research exploring the interpretability of MHE networks. In this paper, we target at exploring how deep neural networks predict height from a single monocular image. Towards a comprehensive understanding of MHE networks, we propose to interpret them from multiple levels: 1) Neurons: unit-level dissection. Exploring the semantic and height selectivity of the lea…
Leveraging synthetically rendered data offers great potential to improve monocular depth estimation and other geometric estimation tasks, but closing the synthetic-real domain gap is a non-trivial and important task. While much recent work has focused on unsupervised domain adaptation, we consider a more realistic scenario where a large amount of synthetic training data is supplemented by a small set of real images with ground-truth. In this setting, we find that existing domain translation approaches are difficult to train and offer little advantage over simple baselines that use a mix of rea…
CCDepth: A Lightweight Self-supervised Depth Estimation Network with Enhanced Interpretability
College of Information Science and Engineering, China University of Petroleum (Beijing)（中国石油大学（北京）信息科学与工程学院）；The University of Hong Kong（香港大学）；Tsinghua University（清华大学）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）
Vision is one of the primary sensing modalities in autonomous driving. In this paper we look at the problem of estimating the velocity of road vehicles from a camera mounted on a moving car. Contrary to prior methods that train end-to-end deep networks that estimate the vehicles' velocity from the video pixels, we propose a two-step approach where first an off-the-shelf tracker is used to extract vehicle bounding boxes and then a small neural network is used to regress the vehicle velocity from the tracked bounding boxes. Surprisingly, we find that this still achieves state-of-the-art estimati…
MVOFormer: Flow-Semantic Transformer for Robust Monocular Visual Odometry
State Key Laboratory of Fluid Power and Mechatronic Systems, Zhejiang University（浙江大学流体动力与机电系统国家重点实验室）；Zhejiang Key Laboratory of Industrial Big Data and Robot Intelligent Systems（浙江省工业大数据与机器人智能系统重点实验室）；School of Mechanical Engineering, Zhejiang University（浙江大学机械工程学院）；Robotics Institute, Zhejiang University（浙江大学机器人研究院）；School of Artificial Intelligence and Robotics, Hunan University（湖南大学人工智能与机器人学院）；Rural Health Research Institute, Charles Sturt University（查尔斯特大学农村健康研究所）；University College London（伦敦大学学院）
Autonomous driving requires efficient reasoning about the location and appearance of the different agents in the scene, which aids in downstream tasks such as object detection, object tracking, and path planning. The past few years have witnessed a surge in approaches that combine the different taskbased modules of the classic self-driving stack into an End-toEnd(E2E) trainable learning system. These approaches replace perception, prediction, and sensor fusion modules with a single contiguous module with shared latent space embedding, from which one extracts a human-interpretable representatio…
GuidedNav: Shaping Vision-Language Navigation Representations through Pre-action Attention and Spatial Guidance
University of Melbourne（墨尔本大学）；AgiBot；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）；University of Electronic Science and Technology of China（电子科技大学）；University of Macau（澳门大学）；University of Technology Sydney（悉尼科技大学）
