Robotic-assisted surgery (RAS) relies on accurate depth estimation for 3D reconstruction and visualization. While foundation models like Depth Anything Models (DAM) show promise, directly applying them to surgery often yields suboptimal results. Fully fine-tuning on limited surgical data can cause overfitting and catastrophic forgetting, compromising model robustness and generalization. Although Low-Rank Adaptation (LoRA) addresses some adaptation issues, its uniform parameter distribution neglects the inherent feature hierarchy, where earlier layers, learning more general features, require more parameters than later ones. To tackle this issue, we introduce Depth Anything in Robotic Endoscopic Surgery (DARES), a novel approach that employs a new adaptation technique, Vector Low-Rank Adaptation (Vector-LoRA) on the DAM V2 to perform self-supervised monocular depth estimation in RAS scenes. To enhance learning efficiency, we introduce Vector-LoRA by integrating more parameters in earlier layers and gradually decreasing parameters in later layers. We also design a reprojection loss based on the multi-scale SSIM error to enhance depth perception by better tailoring the foundation model to the specific requirements of the surgical environment. The proposed method is validated on the SCARED dataset and demonstrates superior performance over recent state-of-the-art self-supervised monocular depth estimation techniques, achieving an improvement of 13.3% in the absolute relative error metric. The code and pre-trained weights are available at https://github.com/mobarakol/DARES.
In this paper, we delve into the realm of 4-D light fields (LFs) to enhance underwater imaging plagued by light absorption, scattering, and other challenges. Contrasting with conventional 2-D RGB imaging, 4-D LF imaging excels in capturing scenes from multiple perspectives, thereby indirectly embedding geometric information. This intrinsic property is anticipated to effectively address the challenges associated with underwater imaging. By leveraging both explicit and implicit depth cues present in 4-D LF images, we propose a progressive, mutually reinforcing framework for underwater 4-D LF image enhancement and depth estimation. Specifically, our framework explicitly utilizes estimated depth information alongside implicit depth-related dynamic convolutional kernels to modulate output features. The entire framework decomposes this complex task, iteratively optimizing the enhanced image and depth information to progressively achieve optimal enhancement results. More importantly, we construct the first 4-D LF-based underwater image dataset for quantitative evaluation and supervised training of learning-based methods, comprising 75 underwater scenes and 3675 high-resolution 2K pairs. To craft vibrant and varied underwater scenes, we build underwater environments with various objects and adopt several types of degradation. Through extensive experimentation, we showcase the potential and superiority of 4-D LF-based underwater imaging vis-a-vis traditional 2-D RGB-based approaches. Moreover, our method effectively corrects color bias and achieves state-of-the-art performance. The dataset and code will be publicly available at https://github.com/linlos1234/LFUIE.
Monocular depth estimation (MDE) and semantic segmentation (SS) are crucial for the navigation and environmental interpretation of many autonomous driving systems. However, their vulnerability to practical adversarial attacks is a significant concern. This paper presents a novel adversarial attack using practical patches that mimic manhole covers to deceive MDE and SS models. The goal is to cause these systems to misinterpret scenes, leading to false detections of near obstacles or non-passable objects. We use Depth Planar Mapping to precisely position these patches on road surfaces, enhancing the attack's effectiveness. Our experiments show that these adversarial patches cause a 43% relative error in MDE and achieve a 96% attack success rate in SS. These patches create affected error regions over twice their size in MDE and approximately equal to their size in SS. Our studies also confirm the patch's effectiveness in physical simulations, the adaptability of the patches across different target models, and the effectiveness of our proposed modules, highlighting their practical implications.
We introduce NimbleD, an efficient self-supervised monocular depth estimation learning framework that incorporates supervision from pseudo-labels generated by a large vision model. This framework does not require camera intrinsics, enabling large-scale pre-training on publicly available videos. Our straightforward yet effective learning strategy significantly enhances the performance of fast and lightweight models without introducing any overhead, allowing them to achieve performance comparable to state-of-the-art self-supervised monocular depth estimation models. This advancement is particularly beneficial for virtual and augmented reality applications requiring low latency inference. The source code, model weights, and acknowledgments are available at https://github.com/xapaxca/nimbled .
Monocular depth estimation in colonoscopy video aims to overcome the unusual lighting properties of the colonoscopic environment. One of the major challenges in this area is the domain gap between annotated but unrealistic synthetic data and unannotated but realistic clinical data. Previous attempts to bridge this domain gap directly target the depth estimation task itself. We propose a general pipeline of structure-preserving synthetic-to-real (sim2real) image translation (producing a modified version of the input image) to retain depth geometry through the translation process. This allows us to generate large quantities of realistic-looking synthetic images for supervised depth estimation with improved generalization to the clinical domain. We also propose a dataset of hand-picked sequences from clinical colonoscopies to improve the image translation process. We demonstrate the simultaneous realism of the translated images and preservation of depth maps via the performance of downstream depth estimation on various datasets.
Hand pose represents key information for action recognition in the egocentric perspective, where the user is interacting with objects. We propose to improve egocentric 3D hand pose estimation based on RGB frames only by using pseudo-depth images. Incorporating state-of-the-art single RGB image depth estimation techniques, we generate pseudo-depth representations of the frames and use distance knowledge to segment irrelevant parts of the scene. The resulting depth maps are then used as segmentation masks for the RGB frames. Experimental results on H2O Dataset confirm the high accuracy of the estimated pose with our method in an action recognition task. The 3D hand pose, together with information from object detection, is processed by a transformer-based action recognition network, resulting in an accuracy of 91.73%, outperforming all state-of-the-art methods. Estimations of 3D hand pose result in competitive performance with existing methods with a mean pose error of 28.66 mm. This method opens up new possibilities for employing distance information in egocentric 3D hand pose estimation without relying on depth sensors.
3D pre-training is crucial to 3D perception tasks. Nevertheless, limited by the difficulties in collecting clean and complete 3D data, 3D pre-training has persistently faced data scaling challenges. In this work, we introduce a novel self-supervised pre-training framework that incorporates millions of images into 3D pre-training corpora by leveraging a large depth estimation model. New pre-training corpora encounter new challenges in representation ability and embedding efficiency of models. Previous pre-training methods rely on farthest point sampling and k-nearest neighbors to embed a fixed number of 3D tokens. However, these approaches prove inadequate when it comes to embedding millions of samples that feature a diverse range of point numbers, spanning from 1,000 to 100,000. In contrast, we propose a tokenizer with linear-time complexity, which enables the efficient embedding of a flexible number of tokens. Accordingly, a new 3D reconstruction target is proposed to cooperate with our 3D tokenizer. Our method achieves state-of-the-art performance in 3D classification, few-shot learning, and 3D segmentation. Code is available at https://github.com/XuechaoChen/P3P-MAE.
Estimating global tree canopy height is crucial for forest conservation and climate change applications. However, capturing high-resolution ground truth canopy height using LiDAR is expensive and not available globally. An efficient alternative is to train a canopy height estimator to operate on single-view remotely sensed imagery. The primary obstacle to this approach is that these methods require significant training data to generalize well globally and across uncommon edge cases. Recent monocular depth estimation foundation models have show strong zero-shot performance even for complex scenes. In this paper we leverage the representations learned by these models to transfer to the remote sensing domain for measuring canopy height. Our findings suggest that our proposed Depth Any Canopy, the result of fine-tuning the Depth Anything v2 model for canopy height estimation, provides a performant and efficient solution, surpassing the current state-of-the-art with superior or comparable performance using only a fraction of the computational resources and parameters. Furthermore, our approach requires less than \$1.30 in compute and results in an estimated carbon footprint of 0.14 kgCO2. Code, experimental results, and model checkpoints are openly available at https://github.com/DarthReca/depth-any-canopy.
Endoscopic surgery relies on two-dimensional views, posing challenges for surgeons in depth perception and instrument manipulation. While Monocular Visual Simultaneous Localization and Mapping (MVSLAM) has emerged as a promising solution, its implementation in endoscopic procedures faces significant challenges due to hardware limitations, such as the use of a monocular camera and the absence of odometry sensors. This study presents BodySLAM, a robust deep learning-based MVSLAM approach that addresses these challenges through three key components: CycleVO, a novel unsupervised monocular pose estimation module; the integration of the state-of-the-art Zoe architecture for monocular depth estimation; and a 3D reconstruction module creating a coherent surgical map. The approach is rigorously evaluated using three publicly available datasets (Hamlyn, EndoSLAM, and SCARED) spanning laparoscopy, gastroscopy, and colonoscopy scenarios, and benchmarked against four state-of-the-art methods. Results demonstrate that CycleVO exhibited competitive performance with the lowest inference time among pose estimation methods, while maintaining robust generalization capabilities, whereas Zoe significantly outperformed existing algorithms for depth estimation in endoscopy. BodySLAM's strong performance across diverse endoscopic scenarios demonstrates its potential as a viable MVSLAM solution for endoscopic applications.
We present Lumina-mGPT, a family of multimodal autoregressive models capable of various vision and language tasks, particularly excelling in generating flexible photorealistic images from text descriptions. By initializing from multimodal Generative PreTraining (mGPT), we demonstrate that decoder-only Autoregressive (AR) model can achieve image generation performance comparable to modern diffusion models with high efficiency through Flexible Progressive Supervised Fine-tuning (FP-SFT). Equipped with our proposed Unambiguous image Representation (UniRep), Lumina-mGPT can flexibly generate high-quality images of varying aspect ratios. Building on the strong image generation capabilities, we further explore Ominiponent Supervised Fine-tuning (Omni-SFT), an initial attempt to elevate Lumina-mGPT into a unified multi-modal generalist. The resulting model demonstrates versatile multimodal capabilities, including visual generation tasks like text-to-image/multiview generation and controllable generation, visual recognition tasks like segmentation and depth estimation, and vision-language tasks like multi-turn visual question answering, showing the rosy potential of the technical direction. Codes and checkpoints are available at https://github.com/Alpha-VLLM/Lumina-mGPT.
We propose HYBRIDDEPTH, a robust depth estimation pipeline that addresses key challenges in depth estimation,including scale ambiguity, hardware heterogeneity, and generalizability. HYBRIDDEPTH leverages focal stack, data conveniently accessible in common mobile devices, to produce accurate metric depth maps. By incorporating depth priors afforded by recent advances in singleimage depth estimation, our model achieves a higher level of structural detail compared to existing methods. We test our pipeline as an end-to-end system, with a newly developed mobile client to capture focal stacks, which are then sent to a GPU-powered server for depth estimation. Comprehensive quantitative and qualitative analyses demonstrate that HYBRIDDEPTH outperforms state-of-the-art(SOTA) models on common datasets such as DDFF12 and NYU Depth V2. HYBRIDDEPTH also shows strong zero-shot generalization. When trained on NYU Depth V2, HYBRIDDEPTH surpasses SOTA models in zero-shot performance on ARKitScenes and delivers more structurally accurate depth maps on Mobile Depth. The code is available at https://github.com/cake-lab/HybridDepth/.
Vision Transformer models trained on large-scale datasets, although effective, often exhibit artifacts in the patch token they extract. While such defects can be alleviated by re-training the entire model with additional classification tokens, the underlying reasons for the presence of these tokens remain unclear. In this paper, we conduct a thorough investigation of this phenomenon, combining theoretical analysis with empirical observations. Our findings reveal that these artifacts originate from the pre-trained network itself, specifically stemming from the leading left singular vector of the network's weights. Furthermore, to mitigate these defects, we propose a novel fine-tuning smooth regularization that rectifies structural deficiencies using only a small dataset, thereby avoiding the need for complete re-training. We validate our method on various downstream tasks, including unsupervised segmentation, classification, supervised segmentation, and depth estimation, demonstrating its effectiveness in improving model performance. Codes and checkpoints are available at https://github.com/haoqiwang/sinder.
We present a novel approach designed to address the complexities posed by challenging, out-of-distribution data in the single-image depth estimation task. Starting with images that facilitate depth prediction due to the absence of unfavorable factors, we systematically generate new, user-defined scenes with a comprehensive set of challenges and associated depth information. This is achieved by leveraging cutting-edge text-to-image diffusion models with depth-aware control, known for synthesizing high-quality image content from textual prompts while preserving the coherence of 3D structure between generated and source imagery. Subsequent fine-tuning of any monocular depth network is carried out through a self-distillation protocol that takes into account images generated using our strategy and its own depth predictions on simple, unchallenging scenes. Experiments on benchmarks tailored for our purposes demonstrate the effectiveness and versatility of our proposal.
Self-supervised monocular depth estimation has gathered notable interest since it can liberate training from dependency on depth annotations. In monocular video training case, recent methods only conduct view synthesis between existing camera views, leading to insufficient guidance. To tackle this, we try to synthesize more virtual camera views by flow-based video frame interpolation (VFI), termed as temporal augmentation. For multi-frame inference, to sidestep the problem of dynamic objects encountered by explicit geometry-based methods like ManyDepth, we return to the feature fusion paradigm and design a VFI-assisted multi-frame fusion module to align and aggregate multi-frame features, using motion and occlusion information obtained by the flow-based VFI model. Finally, we construct a unified self-supervised learning framework, named Mono-ViFI, to bilaterally connect single- and multi-frame depth. In this framework, spatial data augmentation through image affine transformation is incorporated for data diversity, along with a triplet depth consistency loss for regularization. The single- and multi-frame models can share weights, making our framework compact and memory-efficient. Extensive experiments demonstrate that our method can bring significant improvements to current advanced architectures. Source code is available at https://github.com/LiuJF1226/Mono-ViFI.
Large-scale vision models have become integral in many applications due to their unprecedented performance and versatility across downstream tasks. However, the robustness of these foundation models has primarily been explored for a single task, namely image classification. The vulnerability of other common vision tasks, such as semantic segmentation and depth estimation, remains largely unknown. We present a comprehensive empirical evaluation of the adversarial robustness of self-supervised vision encoders across multiple downstream tasks. Our attacks operate in the encoder embedding space and at the downstream task output level. In both cases, current state-of-the-art adversarial fine-tuning techniques tested only for classification significantly degrade clean and robust performance on other tasks. Since the purpose of a foundation model is to cater to multiple applications at once, our findings reveal the need to enhance encoder robustness more broadly. Our code is available at ${github.com/layer6ai-labs/ssl-robustness}$.
Stereo matching provides depth estimation from binocular images for downstream applications. These applications mostly take video streams as input and require temporally consistent depth maps. However, existing methods mainly focus on the estimation at the single-frame level. This commonly leads to temporally inconsistent results, especially in ill-posed regions. In this paper, we aim to leverage temporal information to improve the temporal consistency, accuracy, and efficiency of stereo matching. To achieve this, we formulate video stereo matching as a process of temporal disparity completion followed by continuous iterative refinements. Specifically, we first project the disparity of the previous timestamp to the current viewpoint, obtaining a semi-dense disparity map. Then, we complete this map through a disparity completion module to obtain a well-initialized disparity map. The state features from the current completion module and from the past refinement are fused together, providing a temporally coherent state for subsequent refinement. Based on this coherent state, we introduce a dual-space refinement module to iteratively refine the initialized result in both disparity and disparity gradient spaces, improving estimations in ill-posed regions. Extensive experiments demonstrate that our method effectively alleviates temporal inconsistency while enhancing both accuracy and efficiency.
Recent advancements in camera-based 3D object detection have introduced cross-modal knowledge distillation to bridge the performance gap with LiDAR 3D detectors, leveraging the precise geometric information in LiDAR point clouds. However, existing cross-modal knowledge distillation methods tend to overlook the inherent imperfections of LiDAR, such as the ambiguity of measurements on distant or occluded objects, which should not be transferred to the image detector. To mitigate these imperfections in LiDAR teacher, we propose a novel method that leverages aleatoric uncertainty-free features from ground truth labels. In contrast to conventional label guidance approaches, we approximate the inverse function of the teacher's head to effectively embed label inputs into feature space. This approach provides additional accurate guidance alongside LiDAR teacher, thereby boosting the performance of the image detector. Additionally, we introduce feature partitioning, which effectively transfers knowledge from the teacher modality while preserving the distinctive features of the student, thereby maximizing the potential of both modalities. Experimental results demonstrate that our approach improves mAP and NDS by 5.1 points and 4.9 points compared to the baseline model, proving the effectiveness of our approach. The code is available at https://github.com/sanmin0312/LabelDistill
Accurate depth information is crucial for enhancing the performance of multi-view 3D object detection. Despite the success of some existing multi-view 3D detectors utilizing pixel-wise depth supervision, they overlook two significant phenomena: 1) the depth supervision obtained from LiDAR points is usually distributed on the surface of the object, which is not so friendly to existing DETR-based 3D detectors due to the lack of the depth of 3D object center; 2) for distant objects, fine-grained depth estimation of the whole object is more challenging. Therefore, we argue that the object-wise depth (or 3D center of the object) is essential for accurate detection. In this paper, we propose a new multi-view 3D object detector named OPEN, whose main idea is to effectively inject object-wise depth information into the network through our proposed object-wise position embedding. Specifically, we first employ an object-wise depth encoder, which takes the pixel-wise depth map as a prior, to accurately estimate the object-wise depth. Then, we utilize the proposed object-wise position embedding to encode the object-wise depth information into the transformer decoder, thereby producing 3D object-aware features for final detection. Extensive experiments verify the effectiveness of our proposed method. Furthermore, OPEN achieves a new state-of-the-art performance with 64.4% NDS and 56.7% mAP on the nuScenes test benchmark.
Significant advances have been made in human-centric video generation, yet the joint video-depth generation problem remains underexplored. Most existing monocular depth estimation methods may not generalize well to synthesized images or videos, and multi-view-based methods have difficulty controlling the human appearance and motion. In this work, we present IDOL (unIfied Dual-mOdal Latent diffusion) for high-quality human-centric joint video-depth generation. Our IDOL consists of two novel designs. First, to enable dual-modal generation and maximize the information exchange between video and depth generation, we propose a unified dual-modal U-Net, a parameter-sharing framework for joint video and depth denoising, wherein a modality label guides the denoising target, and cross-modal attention enables the mutual information flow. Second, to ensure a precise video-depth spatial alignment, we propose a motion consistency loss that enforces consistency between the video and depth feature motion fields, leading to harmonized outputs. Additionally, a cross-attention map consistency loss is applied to align the cross-attention map of the video denoising with that of the depth denoising, further facilitating spatial alignment. Extensive experiments on the TikTok and NTU120 datasets show our superior performance, significantly surpassing existing methods in terms of video FVD and depth accuracy.
Self-supervised multi-frame monocular depth estimation relies on the geometric consistency between successive frames under the assumption of a static scene. However, the presence of moving objects in dynamic scenes introduces inevitable inconsistencies, causing misaligned multi-frame feature matching and misleading self-supervision during training. In this paper, we propose a novel framework called ProDepth, which effectively addresses the mismatch problem caused by dynamic objects using a probabilistic approach. We initially deduce the uncertainty associated with static scene assumption by adopting an auxiliary decoder. This decoder analyzes inconsistencies embedded in the cost volume, inferring the probability of areas being dynamic. We then directly rectify the erroneous cost volume for dynamic areas through a Probabilistic Cost Volume Modulation (PCVM) module. Specifically, we derive probability distributions of depth candidates from both single-frame and multi-frame cues, modulating the cost volume by adaptively fusing those distributions based on the inferred uncertainty. Additionally, we present a self-supervision loss reweighting strategy that not only masks out incorrect supervision with high uncertainty but also mitigates the risks in remaining possible dynamic areas in accordance with the probability. Our proposed method excels over state-of-the-art approaches in all metrics on both Cityscapes and KITTI datasets, and demonstrates superior generalization ability on the Waymo Open dataset.