We present SeeingThroughClutter, a method for reconstructing structured 3D representations from single images by segmenting and modeling objects individually. Prior approaches rely on intermediate tasks such as semantic segmentation and depth estimation, which often underperform in complex scenes, particularly in the presence of occlusion and clutter. We address this by introducing an iterative object removal and reconstruction pipeline that decomposes complex scenes into a sequence of simpler subtasks. Using VLMs as orchestrators, foreground objects are removed one at a time via detection, segmentation, object removal, and 3D fitting. We show that removing objects allows for cleaner segmentations of subsequent objects, even in highly occluded scenes. Our method requires no task-specific training and benefits directly from ongoing advances in foundation models. We demonstrate state-of-the-art robustness on 3D-Front and ADE20K datasets. Project Page: https://rioak.github.io/seeingthroughclutter/

We describe a system to remove real-world reflections from images for consumer photography. Our system operates on linear (RAW) photos, and accepts an optional contextual photo looking in the opposite direction (e.g. the "selfie" camera on a mobile device). This optional photo disambiguates what should be considered the reflection. The system is trained solely on synthetic mixtures of real RAW photos, which we combine using a reflection simulation that is photometrically and geometrically accurate. Our system comprises a base model that accepts the captured photo and optional context photo as input, and runs at 256p, followed by an up-sampling model that transforms 256p images to full resolution. The system produces preview images at 1K in 4.5-6.5s on a MacBook or iPhone 14 Pro. We show SOTA results on RAW photos that were captured in the field to embody typical consumer photos, and show that training on RAW simulation data improves performance more than the architectural variations among prior works.

Geometric camera calibration is often required for applications that understand the perspective of the image. We propose Perspective Fields as a representation that models the local perspective properties of an image. Perspective Fields contain per-pixel information about the camera view, parameterized as an Up-vector and a Latitude value. This representation has a number of advantages; it makes minimal assumptions about the camera model and is invariant or equivariant to common image editing operations like cropping, warping, and rotation. It is also more interpretable and aligned with human perception. We train a neural network to predict Perspective Fields and the predicted Perspective Fields can be converted to calibration parameters easily. We demonstrate the robustness of our approach under various scenarios compared with camera calibration-based methods and show example applications in image compositing. Project page: https://jinlinyi.github.io/PerspectiveFields/.

We present the design of a productionized end-to-end stereo depth sensing system that does pre-processing, online stereo rectification, and stereo depth estimation with a fallback to monocular depth estimation when rectification is unreliable. The output of our depth sensing system is then used in a novel view generation pipeline to create 3D computational photography effects using point-of-view images captured by smart glasses. All these steps are executed on-device on the stringent compute budget of a mobile phone, and because we expect the users can use a wide range of smartphones, our design needs to be general and cannot be dependent on a particular hardware or ML accelerator such as a smartphone GPU. Although each of these steps is well studied, a description of a practical system is still lacking. For such a system, all these steps need to work in tandem with one another and fallback gracefully on failures within the system or less than ideal input data. We show how we handle unforeseen changes to calibration, e.g., due to heat, robustly support depth estimation in the wild, and still abide by the memory and latency constraints required for a smooth user experience. We show that our trained models are fast, and run in less than 1s on a six-year-old Samsung Galaxy S8 phone's CPU. Our models generalize well to unseen data and achieve good results on Middlebury and in-the-wild images captured from the smart glasses.

We present an interactive system to ease the creation of so-called video doodles - videos on which artists insert hand-drawn animations for entertainment or educational purposes. Video doodles are challenging to create because to be convincing, the inserted drawings must appear as if they were part of the captured scene. In particular, the drawings should undergo tracking, perspective deformations and occlusions as they move with respect to the camera and to other objects in the scene - visual effects that are difficult to reproduce with existing 2D video editing software. Our system supports these effects by relying on planar canvases that users position in a 3D scene reconstructed from the video. Furthermore, we present a custom tracking algorithm that allows users to anchor canvases to static or dynamic objects in the scene, such that the canvases move and rotate to follow the position and direction of these objects. When testing our system, novices could create a variety of short animated clips in a dozen of minutes, while professionals praised its speed and ease of use compared to existing tools.

We present a novel single-view surface normal estimation method that combines traditional line and vanishing point analysis with a deep learning approach. Starting from a color image and a Manhattan line map, we use a deep neural network to regress on a dense normal map, and a dense Manhattan label map that identifies planar regions aligned with the Manhattan directions. We fuse the normal map and label map in a fully differentiable manner to produce a refined normal map as final output. To do so, we softly decompose the output into a Manhattan part and a non-Manhattan part. The Manhattan part is treated by discrete classification and vanishing points, while the non-Manhattan part is learned by direct supervision. Our method achieves state-of-the-art results on standard single-view normal estimation benchmarks. More importantly, we show that by using vanishing points and lines, our method has better generalization ability than existing works. In addition, we demonstrate how our surface normal network can improve the performance of depth estimation networks, both quantitatively and qualitatively, in particular, in 3D reconstructions of walls and other flat surfaces.

3D photography is a new medium that allows viewers to more fully experience a captured moment. In this work, we refer to a 3D photo as one that displays parallax induced by moving the viewpoint (as opposed to a stereo pair with a fixed viewpoint). 3D photos are static in time, like traditional photos, but are displayed with interactive parallax on mobile or desktop screens, as well as on Virtual Reality devices, where viewing it also includes stereo. We present an end-to-end system for creating and viewing 3D photos, and the algorithmic and design choices therein. Our 3D photos are captured in a single shot and processed directly on a mobile device. The method starts by estimating depth from the 2D input image using a new monocular depth estimation network that is optimized for mobile devices. It performs competitively to the state-of-the-art, but has lower latency and peak memory consumption and uses an order of magnitude fewer parameters. The resulting depth is lifted to a layered depth image, and new geometry is synthesized in parallax regions. We synthesize color texture and structures in the parallax regions as well, using an inpainting network, also optimized for mobile devices, on the LDI directly. Finally, we convert the result into a mesh-based representation that can be efficiently transmitted and rendered even on low-end devices and over poor network connections. Altogether, the processing takes just a few seconds on a mobile device, and the result can be instantly viewed and shared. We perform extensive quantitative evaluation to validate our system and compare its new components against the current state-of-the-art.

We present an algorithm for reconstructing dense, geometrically consistent depth for all pixels in a monocular video. We leverage a conventional structure-from-motion reconstruction to establish geometric constraints on pixels in the video. Unlike the ad-hoc priors in classical reconstruction, we use a learning-based prior, i.e., a convolutional neural network trained for single-image depth estimation. At test time, we fine-tune this network to satisfy the geometric constraints of a particular input video, while retaining its ability to synthesize plausible depth details in parts of the video that are less constrained. We show through quantitative validation that our method achieves higher accuracy and a higher degree of geometric consistency than previous monocular reconstruction methods. Visually, our results appear more stable. Our algorithm is able to handle challenging hand-held captured input videos with a moderate degree of dynamic motion. The improved quality of the reconstruction enables several applications, such as scene reconstruction and advanced video-based visual effects.

Understanding fashion styles and trends is of great potential interest to retailers and consumers alike. The photos people upload to social media are a historical and public data source of how people dress across the world and at different times. While we now have tools to automatically recognize the clothing and style attributes of what people are wearing in these photographs, we lack the ability to analyze spatial and temporal trends in these attributes or make predictions about the future. In this paper we address this need by providing an automatic framework that analyzes large corpora of street imagery to (a) discover and forecast long-term trends of various fashion attributes as well as automatically discovered styles, and (b) identify spatio-temporally localized events that affect what people wear. We show that our framework makes long term trend forecasts that are > 20% more accurate than prior art, and identifies hundreds of socially meaningful events that impact fashion across the globe.

In cinema, large camera lenses create beautiful shallow depth of field (DOF), but make focusing difficult and expensive. Accurate cinema focus usually relies on a script and a person to control focus in realtime. Casual videographers often crave cinematic focus, but fail to achieve it. We either sacrifice shallow DOF, as in smartphone videos; or we struggle to deliver accurate focus, as in videos from larger cameras. This paper is about a new approach in the pursuit of cinematic focus for casual videography. We present a system that synthetically renders refocusable video from a deep DOF video shot with a smartphone, and analyzes future video frames to deliver context-aware autofocus for the current frame. To create refocusable video, we extend recent machine learning methods designed for still photography, contributing a new dataset for machine training, a rendering model better suited to cinema focus, and a filtering solution for temporal coherence. To choose focus accurately for each frame, we demonstrate autofocus that looks at upcoming video frames and applies AI-assist modules such as motion, face, audio and saliency detection. We also show that autofocus benefits from machine learning and a large-scale video dataset with focus annotation, where we use our RVR-LAAF GUI to create this sizable dataset efficiently. We deliver, for example, a shallow DOF video where the autofocus transitions onto each person before she begins to speak. This is impossible for conventional camera autofocus because it would require seeing into the future.

We present DeepMVS, a deep convolutional neural network (ConvNet) for multi-view stereo reconstruction. Taking an arbitrary number of posed images as input, we first produce a set of plane-sweep volumes and use the proposed DeepMVS network to predict high-quality disparity maps. The key contributions that enable these results are (1) supervised pretraining on a photorealistic synthetic dataset, (2) an effective method for aggregating information across a set of unordered images, and (3) integrating multi-layer feature activations from the pre-trained VGG-19 network. We validate the efficacy of DeepMVS using the ETH3D Benchmark. Our results show that DeepMVS compares favorably against state-of-the-art conventional MVS algorithms and other ConvNet based methods, particularly for near-textureless regions and thin structures.

A number of consumer-grade spherical cameras have recently appeared, enabling affordable monoscopic VR content creation in the form of full 360° X 180° spherical panoramic photos and videos. While monoscopic content is certainly engaging, it fails to leverage a main aspect of VR HMDs, namely stereoscopic display. Recent stereoscopic capture rigs involve placing many cameras in a ring and synthesizing an omni-directional stereo panorama enabling a user to look around to explore the scene in stereo. In this work, we describe a method that takes images from two 360° spherical cameras and synthesizes an omni-directional stereo panorama with stereo in all directions. Our proposed method has a lower equipment cost than camera-ring alternatives, can be assembled with currently available off-the-shelf equipment, and is relatively small and light-weight compared to the alternatives. We validate our method by generating both stills and videos. We have conducted a user study to better understand what kinds of geometric processing are necessary for a pleasant viewing experience. We also discuss several algorithmic variations, each with their own time and quality trade-offs.

We propose a new method for turning an Internet-scale corpus of categorized images into a small set of human-interpretable discriminative visual elements using powerful tools based on deep learning. A key challenge with deep learning methods is generating human-interpretable models. To address this, we propose a new technique that uses bubble images -- images where most of the content has been obscured -- to identify spatially localized, discriminative content in each image. By modifying the model training procedure to use both the source imagery and these bubble images, we can arrive at final models which retain much of the original classification performance, but are much more amenable to identifying interpretable visual elements. We apply our algorithm to a wide variety of datasets, including two new Internet-scale datasets of people and places, and show applications to visual mining and discovery. Our method is simple, scalable, and produces visual elements that are highly representative compared to prior work.

Geometry and geography can play an important role in recognition tasks in computer vision. To aid in studying connections between geometry and recognition, we introduce NYC3DCars, a rich dataset for vehicle detection in urban scenes built from Internet photos drawn from the wild, focused on densely trafficked areas of New York City. Our dataset is augmented with detailed geometric and geographic information, including full camera poses derived from structure from motion, 3D vehicle annotations, and geographic information from open resources, including road segmentations and directions of travel. NYC3DCars can be used to study new questions about using geometric information in detection tasks, and to explore applications of Internet photos in understanding cities. To demonstrate the utility of our data, we evaluate the use of the geographic information in our dataset to enhance a parts-based detection method, and suggest other avenues for future exploration.