Event cameras capture scene changes asynchronously on a per-pixel basis, enabling extremely high temporal resolution. However, this advantage comes at the cost of high bandwidth, memory, and computational demands. To address this, prior work has explored event downsampling, but most approaches rely on fixed heuristics or threshold-based strategies, limiting their adaptability. Instead, we propose a probabilistic framework, POLED, that models event importance through an event-importance probability density function (ePDF), which can be arbitrarily defined and adapted to different applications. Our approach operates in a purely online setting, estimating event importance on-the-fly from raw event streams, enabling scene-specific adaptation. Additionally, we introduce zero-shot event downsampling, where downsampled events must remain usable for models trained on the original event stream, without task-specific adaptation. We design a contour-preserving ePDF that prioritizes structurally important events and evaluate our method across four datasets and tasks--object classification, image interpolation, surface normal estimation, and object detection--demonstrating that intelligent sampling is crucial for maintaining performance under event-budget constraints.
This work introduces a novel method for surface normal estimation from rectified stereo image pairs, leveraging affine transformations derived from disparity values to achieve fast and accurate results. We demonstrate how the rectification of stereo image pairs simplifies the process of surface normal estimation by reducing computational complexity. To address noise reduction, we develop a custom algorithm inspired by convolutional operations, tailored to process disparity data efficiently. We also introduce adaptive heuristic techniques for efficiently detecting connected surface components within the images, further improving the robustness of the method. By integrating these methods, we construct a surface normal estimator that is both fast and accurate, producing a dense, oriented point cloud as the final output. Our method is validated using both simulated environments and real-world stereo images from the Middlebury and Cityscapes datasets, demonstrating significant improvements in real-time performance and accuracy when implemented on a GPU. Upon acceptance, the shader source code will be made publicly available to facilitate further research and reproducibility.
Surface normal estimation serves as a cornerstone for a spectrum of computer vision applications. While numerous efforts have been devoted to static image scenarios, ensuring temporal coherence in video-based normal estimation remains a formidable challenge. Instead of merely augmenting existing methods with temporal components, we present NormalCrafter to leverage the inherent temporal priors of video diffusion models. To secure high-fidelity normal estimation across sequences, we propose Semantic Feature Regularization (SFR), which aligns diffusion features with semantic cues, encouraging the model to concentrate on the intrinsic semantics of the scene. Moreover, we introduce a two-stage training protocol that leverages both latent and pixel space learning to preserve spatial accuracy while maintaining long temporal context. Extensive evaluations demonstrate the efficacy of our method, showcasing a superior performance in generating temporally consistent normal sequences with intricate details from diverse videos.
While surface normals are widely used to analyse 3D scene geometry, surface normal estimation from LiDAR point clouds remains severely underexplored. This is caused by the lack of large-scale annotated datasets on the one hand, and lack of methods that can robustly handle the sparse and often noisy LiDAR data in a reasonable time on the other hand. We address these limitations using a traffic simulation engine and present LiSu, the first large-scale, synthetic LiDAR point cloud dataset with ground truth surface normal annotations, eliminating the need for tedious manual labeling. Additionally, we propose a novel method that exploits the spatiotemporal characteristics of autonomous driving data to enhance surface normal estimation accuracy. By incorporating two regularization terms, we enforce spatial consistency among neighboring points and temporal smoothness across consecutive LiDAR frames. These regularizers are particularly effective in self-training settings, where they mitigate the impact of noisy pseudo-labels, enabling robust real-world deployment. We demonstrate the effectiveness of our method on LiSu, achieving state-of-the-art performance in LiDAR surface normal estimation. Moreover, we showcase its full potential in addressing the challenging task of synthetic-to-real domain adaptation, leading to improved neural surface reconstruction on real-world data.
Photometric stereo (PS) endeavors to ascertain surface normals using shading clues from photometric images under various illuminations. Recent deep learning-based PS methods often overlook the complexity of object surfaces. These neural network models, which exclusively rely on photometric images for training, often produce blurred results in high-frequency regions characterized by local discontinuities, such as wrinkles and edges with significant gradient changes. To address this, we propose the Image Gradient-Aided Photometric Stereo Network (IGA-PSN), a dual-branch framework extracting features from both photometric images and their gradients. Furthermore, we incorporate an hourglass regression network along with supervision to regularize normal regression. Experiments on DiLiGenT benchmarks show that IGA-PSN outperforms previous methods in surface normal estimation, achieving a mean angular error of 6.46 while preserving textures and geometric shapes in complex regions.
Geometric estimation is required for scene understanding and analysis in panoramic 360{\deg} images. Current methods usually predict a single feature, such as depth or surface normal. These methods can lack robustness, especially when dealing with intricate textures or complex object surfaces. We introduce a novel multi-task learning (MTL) network that simultaneously estimates depth and surface normals from 360{\deg} images. Our first innovation is our MTL architecture, which enhances predictions for both tasks by integrating geometric information from depth and surface normal estimation, enabling a deeper understanding of 3D scene structure. Another innovation is our fusion module, which bridges the two tasks, allowing the network to learn shared representations that improve accuracy and robustness. Experimental results demonstrate that our MTL architecture significantly outperforms state-of-the-art methods in both depth and surface normal estimation, showing superior performance in complex and diverse scenes. Our model's effectiveness and generalizability, particularly in handling intricate surface textures, establish it as a new benchmark in 360{\deg} image geometric estimation. The code and model are available at \url{https://github.com/huangkun101230/360MTLGeometricEstimation}.
Recent work showed that large diffusion models can be reused as highly precise monocular depth estimators by casting depth estimation as an image-conditional image generation task. While the proposed model achieved state-of-the-art results, high computational demands due to multi-step inference limited its use in many scenarios. In this paper, we show that the perceived inefficiency was caused by a flaw in the inference pipeline that has so far gone unnoticed. The fixed model performs comparably to the best previously reported configuration while being more than 200$\times$ faster. To optimize for downstream task performance, we perform end-to-end fine-tuning on top of the single-step model with task-specific losses and get a deterministic model that outperforms all other diffusion-based depth and normal estimation models on common zero-shot benchmarks. We surprisingly find that this fine-tuning protocol also works directly on Stable Diffusion and achieves comparable performance to current state-of-the-art diffusion-based depth and normal estimation models, calling into question some of the conclusions drawn from prior works.
We present Sapiens, a family of models for four fundamental human-centric vision tasks -- 2D pose estimation, body-part segmentation, depth estimation, and surface normal prediction. Our models natively support 1K high-resolution inference and are extremely easy to adapt for individual tasks by simply fine-tuning models pretrained on over 300 million in-the-wild human images. We observe that, given the same computational budget, self-supervised pretraining on a curated dataset of human images significantly boosts the performance for a diverse set of human-centric tasks. The resulting models exhibit remarkable generalization to in-the-wild data, even when labeled data is scarce or entirely synthetic. Our simple model design also brings scalability -- model performance across tasks improves as we scale the number of parameters from 0.3 to 2 billion. Sapiens consistently surpasses existing baselines across various human-centric benchmarks. We achieve significant improvements over the prior state-of-the-art on Humans-5K (pose) by 7.6 mAP, Humans-2K (part-seg) by 17.1 mIoU, Hi4D (depth) by 22.4% relative RMSE, and THuman2 (normal) by 53.5% relative angular error. Project page: https://about.meta.com/realitylabs/codecavatars/sapiens.
360 cameras capture the entire surrounding environment with a large FoV, exhibiting comprehensive visual information to directly infer the 3D structures, e.g., depth and surface normal, and semantic information simultaneously. Existing works predominantly specialize in a single task, leaving multi-task learning of 3D geometry and semantics largely unexplored. Achieving such an objective is, however, challenging due to: 1) inherent spherical distortion of planar equirectangular projection (ERP) and insufficient global perception induced by 360 image's ultra-wide FoV; 2) non-trivial progress in effectively merging geometry and semantics among different tasks to achieve mutual benefits. In this paper, we propose a novel end-to-end multi-task learning framework, named Elite360M, capable of inferring 3D structures via depth and surface normal estimation, and semantics via semantic segmentation simultaneously. Our key idea is to build a representation with strong global perception and less distortion while exploring the inter- and cross-task relationships between geometry and semantics. We incorporate the distortion-free and spatially continuous icosahedron projection (ICOSAP) points and combine them with ERP to enhance global perception. With a negligible cost, a Bi-projection Bi-attention Fusion module is thus designed to capture the semantic- and distance-aware dependencies between each pixel of the region-aware ERP feature and the ICOSAP point feature set. Moreover, we propose a novel Cross-task Collaboration module to explicitly extract task-specific geometric and semantic information from the learned representation to achieve preliminary predictions. It then integrates the spatial contextual information among tasks to realize cross-task fusion. Extensive experiments demonstrate the effectiveness and efficacy of Elite360M.
This work addresses the challenge of high-quality surface normal estimation from monocular colored inputs (i.e., images and videos), a field which has recently been revolutionized by repurposing diffusion priors. However, previous attempts still struggle with stochastic inference, conflicting with the deterministic nature of the Image2Normal task, and costly ensembling step, which slows down the estimation process. Our method, StableNormal, mitigates the stochasticity of the diffusion process by reducing inference variance, thus producing "Stable-and-Sharp" normal estimates without any additional ensembling process. StableNormal works robustly under challenging imaging conditions, such as extreme lighting, blurring, and low quality. It is also robust against transparent and reflective surfaces, as well as cluttered scenes with numerous objects. Specifically, StableNormal employs a coarse-to-fine strategy, which starts with a one-step normal estimator (YOSO) to derive an initial normal guess, that is relatively coarse but reliable, then followed by a semantic-guided refinement process (SG-DRN) that refines the normals to recover geometric details. The effectiveness of StableNormal is demonstrated through competitive performance in standard datasets such as DIODE-indoor, iBims, ScannetV2 and NYUv2, and also in various downstream tasks, such as surface reconstruction and normal enhancement. These results evidence that StableNormal retains both the "stability" and "sharpness" for accurate normal estimation. StableNormal represents a baby attempt to repurpose diffusion priors for deterministic estimation. To democratize this, code and models have been publicly available in hf.co/Stable-X
This paper introduces PanoRadar, a novel RF imaging system that brings RF resolution close to that of LiDAR, while providing resilience against conditions challenging for optical signals. Our LiDAR-comparable 3D imaging results enable, for the first time, a variety of visual recognition tasks at radio frequency, including surface normal estimation, semantic segmentation, and object detection. PanoRadar utilizes a rotating single-chip mmWave radar, along with a combination of novel signal processing and machine learning algorithms, to create high-resolution 3D images of the surroundings. Our system accurately estimates robot motion, allowing for coherent imaging through a dense grid of synthetic antennas. It also exploits the high azimuth resolution to enhance elevation resolution using learning-based methods. Furthermore, PanoRadar tackles 3D learning via 2D convolutions and addresses challenges due to the unique characteristics of RF signals. Our results demonstrate PanoRadar's robust performance across 12 buildings.
The presence of spherical distortion on the Equirectangular image is an acknowledged challenge in dense regression computer vision tasks, such as surface normal estimation. Recent advances in convolutional neural networks (CNNs) strive to mitigate spherical distortion but often fall short in capturing holistic structures effectively, primarily due to their fixed receptive field. On the other hand, vision transformers (ViTs) excel in establishing long-range dependencies through a global self-attention mechanism, yet they encounter limitations in preserving local details. We introduce \textit{PanoNormal}, a monocular surface normal estimation architecture designed for 360{\deg} images, which combines the strengths of CNNs and ViTs. Specifically, we employ a multi-level global self-attention scheme with the consideration of the spherical feature distribution, enhancing the comprehensive understanding of the scene. Our experimental results demonstrate that our approach achieves state-of-the-art performance across multiple popular 360{\deg} monocular datasets. The code and models will be released.
We introduce Metric3D v2, a geometric foundation model for zero-shot metric depth and surface normal estimation from a single image, which is crucial for metric 3D recovery. While depth and normal are geometrically related and highly complimentary, they present distinct challenges. SoTA monocular depth methods achieve zero-shot generalization by learning affine-invariant depths, which cannot recover real-world metrics. Meanwhile, SoTA normal estimation methods have limited zero-shot performance due to the lack of large-scale labeled data. To tackle these issues, we propose solutions for both metric depth estimation and surface normal estimation. For metric depth estimation, we show that the key to a zero-shot single-view model lies in resolving the metric ambiguity from various camera models and large-scale data training. We propose a canonical camera space transformation module, which explicitly addresses the ambiguity problem and can be effortlessly plugged into existing monocular models. For surface normal estimation, we propose a joint depth-normal optimization module to distill diverse data knowledge from metric depth, enabling normal estimators to learn beyond normal labels. Equipped with these modules, our depth-normal models can be stably trained with over 16 million of images from thousands of camera models with different-type annotations, resulting in zero-shot generalization to in-the-wild images with unseen camera settings. Our method enables the accurate recovery of metric 3D structures on randomly collected internet images, paving the way for plausible single-image metrology. Our project page is at https://JUGGHM.github.io/Metric3Dv2.
In this study, we address the intricate challenge of multi-task dense prediction, encompassing tasks such as semantic segmentation, depth estimation, and surface normal estimation, particularly when dealing with partially annotated data (MTPSL). The complexity arises from the absence of complete task labels for each training image. Given the inter-related nature of these pixel-wise dense tasks, our focus is on mining and capturing cross-task relationships. Existing solutions typically rely on learning global image representations for global cross-task image matching, imposing constraints that, unfortunately, sacrifice the finer structures within the images. Attempting local matching as a remedy faces hurdles due to the lack of precise region supervision, making local alignment a challenging endeavor. The introduction of Segment Anything Model (SAM) sheds light on addressing local alignment challenges by providing free and high-quality solutions for region detection. Leveraging SAM-detected regions, the subsequent challenge lies in aligning the representations within these regions. Diverging from conventional methods that directly learn a monolithic image representation, our proposal involves modeling region-wise representations using Gaussian Distributions. Aligning these distributions between corresponding regions from different tasks imparts higher flexibility and capacity to capture intra-region structures, accommodating a broader range of tasks. This innovative approach significantly enhances our ability to effectively capture cross-task relationships, resulting in improved overall performance in partially supervised multi-task dense prediction scenarios. Extensive experiments conducted on two widely used benchmarks underscore the superior effectiveness of our proposed method, showcasing state-of-the-art performance even when compared to fully supervised methods.
Extensive pre-training with large data is indispensable for downstream geometry and semantic visual perception tasks. Thanks to large-scale text-to-image (T2I) pretraining, recent works show promising results by simply fine-tuning T2I diffusion models for dense perception tasks. However, several crucial design decisions in this process still lack comprehensive justification, encompassing the necessity of the multi-step stochastic diffusion mechanism, training strategy, inference ensemble strategy, and fine-tuning data quality. In this work, we conduct a thorough investigation into critical factors that affect transfer efficiency and performance when using diffusion priors. Our key findings are: 1) High-quality fine-tuning data is paramount for both semantic and geometry perception tasks. 2) The stochastic nature of diffusion models has a slightly negative impact on deterministic visual perception tasks. 3) Apart from fine-tuning the diffusion model with only latent space supervision, task-specific image-level supervision is beneficial to enhance fine-grained details. These observations culminate in the development of GenPercept, an effective deterministic one-step fine-tuning paradigm tailed for dense visual perception tasks. Different from the previous multi-step methods, our paradigm has a much faster inference speed, and can be seamlessly integrated with customized perception decoders and loss functions for image-level supervision, which is critical to improving the fine-grained details of predictions. Comprehensive experiments on diverse dense visual perceptual tasks, including monocular depth estimation, surface normal estimation, image segmentation, and matting, are performed to demonstrate the remarkable adaptability and effectiveness of our proposed method.
Despite the growing demand for accurate surface normal estimation models, existing methods use general-purpose dense prediction models, adopting the same inductive biases as other tasks. In this paper, we discuss the inductive biases needed for surface normal estimation and propose to (1) utilize the per-pixel ray direction and (2) encode the relationship between neighboring surface normals by learning their relative rotation. The proposed method can generate crisp - yet, piecewise smooth - predictions for challenging in-the-wild images of arbitrary resolution and aspect ratio. Compared to a recent ViT-based state-of-the-art model, our method shows a stronger generalization ability, despite being trained on an orders of magnitude smaller dataset. The code is available at https://github.com/baegwangbin/DSINE.
We propose the use of a Transformer to accurately predict normals from point clouds with noise and density variations. Previous learning-based methods utilize PointNet variants to explicitly extract multi-scale features at different input scales, then focus on a surface fitting method by which local point cloud neighborhoods are fitted to a geometric surface approximated by either a polynomial function or a multi-layer perceptron (MLP). However, fitting surfaces to fixed-order polynomial functions can suffer from overfitting or underfitting, and learning MLP-represented hyper-surfaces requires pre-generated per-point weights. To avoid these limitations, we first unify the design choices in previous works and then propose a simplified Transformer-based model to extract richer and more robust geometric features for the surface normal estimation task. Through extensive experiments, we demonstrate that our Transformer-based method achieves state-of-the-art performance on both the synthetic shape dataset PCPNet, and the real-world indoor scene dataset SceneNN, exhibiting more noise-resilient behavior and significantly faster inference. Most importantly, we demonstrate that the sophisticated hand-designed modules in existing works are not necessary to excel at the task of surface normal estimation.
Recent advances in event-based shape determination from polarization offer a transformative approach that tackles the trade-off between speed and accuracy in capturing surface geometries. In this paper, we investigate event-based shape from polarization using Spiking Neural Networks (SNNs), introducing the Single-Timestep and Multi-Timestep Spiking UNets for effective and efficient surface normal estimation. Specificially, the Single-Timestep model processes event-based shape as a non-temporal task, updating the membrane potential of each spiking neuron only once, thereby reducing computational and energy demands. In contrast, the Multi-Timestep model exploits temporal dynamics for enhanced data extraction. Extensive evaluations on synthetic and real-world datasets demonstrate that our models match the performance of state-of-the-art Artifical Neural Networks (ANNs) in estimating surface normals, with the added advantage of superior energy efficiency. Our work not only contributes to the advancement of SNNs in event-based sensing but also sets the stage for future explorations in optimizing SNN architectures, integrating multi-modal data, and scaling for applications on neuromorphic hardware.
Transparent objects are widely used in our daily lives, making it important to teach robots to interact with them. However, it's not easy because the reflective and refractive effects can make depth cameras fail to give accurate geometry measurements. To solve this problem, this paper introduces RFTrans, an RGB-D-based method for surface normal estimation and manipulation of transparent objects. By leveraging refractive flow as an intermediate representation, the proposed method circumvents the drawbacks of directly predicting the geometry (e.g. surface normal) from images and helps bridge the sim-to-real gap. It integrates the RFNet, which predicts refractive flow, object mask, and boundaries, followed by the F2Net, which estimates surface normal from the refractive flow. To make manipulation possible, a global optimization module will take in the predictions, refine the raw depth, and construct the point cloud with normal. An off-the-shelf analytical grasp planning algorithm is followed to generate the grasp poses. We build a synthetic dataset with physically plausible ray-tracing rendering techniques to train the networks. Results show that the proposed method trained on the synthetic dataset can consistently outperform the baseline method in both synthetic and real-world benchmarks by a large margin. Finally, a real-world robot grasping task witnesses an 83% success rate, proving that refractive flow can help enable direct sim-to-real transfer. The code, data, and supplementary materials are available at https://rftrans.robotflow.ai.
Dense prediction tasks, such as semantic segmentation, depth estimation, and surface normal prediction, can be easily formulated as per-pixel classification (discrete outputs) or regression (continuous outputs). This per-pixel prediction paradigm has remained popular due to the prevalence of fully convolutional networks. However, on the recent frontier of segmentation task, the community has been witnessing a shift of paradigm from per-pixel prediction to cluster-prediction with the emergence of transformer architectures, particularly the mask transformers, which directly predicts a label for a mask instead of a pixel. Despite this shift, methods based on the per-pixel prediction paradigm still dominate the benchmarks on the other dense prediction tasks that require continuous outputs, such as depth estimation and surface normal prediction. Motivated by the success of DORN and AdaBins in depth estimation, achieved by discretizing the continuous output space, we propose to generalize the cluster-prediction based method to general dense prediction tasks. This allows us to unify dense prediction tasks with the mask transformer framework. Remarkably, the resulting model PolyMaX demonstrates state-of-the-art performance on three benchmarks of NYUD-v2 dataset. We hope our simple yet effective design can inspire more research on exploiting mask transformers for more dense prediction tasks. Code and model will be made available.