CVPR 2025

The dream of “Holographic” communication—where you can view a remote event from any angle in real-time—has long been a staple of science fiction. In computer vision, this is known as Free-Viewpoint Video (FVV). The goal is to reconstruct dynamic 3D scenes from multiple camera feeds instantly. While recent technologies like 3D Gaussian Splatting (3DGS) have revolutionized how we render static scenes, handling dynamic scenes (videos where people and objects move) remains a massive computational bottleneck. Traditional methods require processing the entire video offline, which is useless for live interactions like virtual meetings or sports broadcasting. Even existing “streaming” methods often take over 10 seconds to process a single frame, creating an unacceptable lag. ...

Imagine asking a robot to “pick up the red mug next to the laptop.” To us, this is a trivial request. To an AI, it is a geometric and semantic nightmare. The AI must identify objects in 3D space, understand what “red” and “mug” look like, and figure out the spatial relationship “next to.” While Large Language Models (LLMs) have mastered text, and Vision-Language Models (VLMs) have conquered 2D images, 3D scene understanding remains a frontier filled with challenges. Most current approaches awkwardly stitch together 2D image data and 3D point clouds, often losing the fine-grained details that make a scene coherent. They struggle to understand how objects relate to one another in physical space and are notoriously computationally expensive. ...

Introduction In the rapid evolution of Artificial Intelligence, we have witnessed a shift from models that simply predict the next word to models that can solve complex logic puzzles. With the release of systems like OpenAI’s o1, text-based Large Language Models (LLMs) have demonstrated “System 2” thinking—the ability to deliberate, reason step-by-step, and self-correct before answering. However, there is a glaring gap in this progress: Vision. While Multimodal Large Language Models (MLLMs)—models that can see and talk—have become excellent at describing images (perception), they often struggle when asked to perform complex reasoning about what they see. If you show an AI a chart and ask for a deep economic analysis, or a geometric figure and ask for a multi-step proof, it frequently hallucinates or takes a shortcut to the wrong answer. ...

Imagine a drone flying at high speed through a dimly lit tunnel. A standard camera would likely fail in this scenario; the fast motion causes severe blur, and the low light results in grainy, unusable footage. This is the bottleneck for many robotics applications today. However, there is a different kind of sensor that thrives in exactly these conditions: the Event Camera. Event cameras are bio-inspired sensors that work differently than the cameras in our phones. Instead of capturing full frames at a fixed rate, they operate asynchronously, detecting changes in brightness at the pixel level. This gives them incredible advantages: microsecond latency, high dynamic range, and no motion blur. ...

If you have ever played with Text-to-Image (T2I) models like Stable Diffusion, you know the struggle: you type a prompt, get a weird result, tweak the prompt, get a slightly less weird result, and repeat. While these models are powerful, they aren’t naturally aligned with human aesthetic preferences or detailed instruction following. In the world of Large Language Models (LLMs), we solved this using Reinforcement Learning from Human Feedback (RLHF) and, more recently, Direct Preference Optimization (DPO). These methods take “winning” and “losing” outputs and teach the model to prefer the winner. ...

Imagine you have a digital photo album containing thousands of images. You want to find a picture of your specific pet, “Fido,” catching a frisbee. You type “Fido catching a frisbee” into the search bar. Standard Vision-Language Models (VLMs) like CLIP differ from standard object detectors because they can understand open-ended text. However, they have a major limitation: they know what a dog looks like, but they don’t know what your dog, Fido, looks like. ...

The arrival of 3D Gaussian Splatting (3DGS) marked a paradigm shift in neural rendering. Unlike Neural Radiance Fields (NeRFs), which rely on expensive ray marching through implicit volumes, 3DGS utilizes explicit point clouds—specifically, 3D Gaussians—to render scenes in real-time with photorealistic quality. However, despite its speed and visual fidelity, 3DGS has a “messy room” problem. The quality of the final render is heavily dependent on how the Gaussian points are distributed. If the initialization (usually via Structure from Motion) is poor, or if the optimization process fails to place points where they are needed, the model suffers from artifacts. You might see “floaters” (random blobs floating in space), blurred details in complex geometry, or erroneous depth estimations. ...

Beyond Matching: How Implicit Learning Solves Image-to-Point Cloud Registration Imagine you are a robot navigating a city. You have a pre-built 3D map of the city (a point cloud), and you just took a picture with your onboard camera. To know where you are, you need to figure out exactly where that 2D picture fits inside that massive 3D world. This problem is known as Image-to-Point Cloud Registration. It sounds simple in theory—just line up the picture with the 3D model—but in practice, it is incredibly difficult. Why? Because you are trying to match two completely different types of data: a 2D grid of pixels (the image) and an unordered set of spatial coordinates (the point cloud). ...

Dreaming of Data: How ImagineFSL Revolutionizes Few-Shot Learning with Synthetic Pretraining In the world of deep learning, data is the fuel that powers the engine. But what happens when that fuel runs low? This is the core challenge of Few-Shot Learning (FSL)—teaching a model to recognize new concepts with only one or a handful of examples. Recently, Vision-Language Models (VLMs) like CLIP have shown incredible promise in this area. However, adapting these massive models to specific tasks with tiny datasets remains a hurdle. The community’s recent answer has been Generative AI. If we don’t have enough data, why not just generate it using Text-to-Image (T2I) models like Stable Diffusion? ...

Solving the Ghosting Problem: How to Reconstruct Images from Multiplexed Single-Photon Detectors Imagine trying to take a picture in near-total darkness, where light is so scarce that you are counting individual particles—photons—as they hit the sensor. This is the realm of single-photon detectors. These devices are revolutionizing fields ranging from biological imaging and Lidar to quantum optics. Among these, Superconducting Nanowire Single-Photon Detectors (SNSPDs) are the gold standard. They offer incredible efficiency and precision. However, they have a major scaling problem: they need to be cooled to cryogenic temperatures. If you want a megapixel camera, you can’t easily run a million wires out of a deep-freeze cryostat without introducing too much heat. ...

Have you ever compressed an image so much that it looked blocky and pixelated, yet your phone still recognized the face in it perfectly? Conversely, have you ever taken a photo that looked fine to you, but your smart camera refused to focus or detect the object? For decades, the field of image processing has been obsessed with one question: “Does this look good to a human?” We built compression algorithms (like JPEG), cameras, and restoration filters designed to please the Human Visual System (HVS). But the world has changed. According to recent data, Machine-to-Machine (M2M) connections have surpassed Human-to-Machine connections. Today, the primary consumer of visual data isn’t you or me—it’s Artificial Intelligence. ...

The dream of Virtual Reality (VR) has always been the “Holodeck” concept—the ability to step into a digital recording of the real world and experience it exactly as if you were there. You want to be able to walk around, lean in to see details, look behind you, and hear the soundscape change as you move. While we have made massive strides with technologies like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting, we hit a wall when it comes to dynamic scenes. Most current datasets are either static (frozen in time), object-centric (looking at a single object from the outside), or silent (no audio). ...

Imagine you want to direct a short film. You have a script, and you have a photo of your lead actor. In the traditional world, this requires cameras, lighting crews, and days of shooting. In the world of Generative AI, we have moved closer to doing this with just a text prompt. However, if you have ever tried to generate a video of a specific person using standard text-to-video models, you have likely encountered the “Identity Problem.” You might upload a photo of yourself and ask for a video of you playing basketball. The result? A person playing basketball who looks vaguely like you in one frame, like your cousin in the next, and like a complete stranger in the third. ...

If you have ever tried to run a state-of-the-art Large Language Model (LLM) like Llama-2 or a vision model like Segment Anything (SAM) on a single consumer-grade GPU, you know the struggle. These models are massive. A 7-billion parameter model is often the upper limit of what a decent desktop GPU can handle for inference, let alone fine-tuning. To deploy these models efficiently, we often turn to pruning—the process of removing unnecessary weights to make the model smaller and faster. However, there is a catch. Current “one-shot” pruning methods (which are fast and don’t require expensive retraining) work great when you remove 20% or 30% of the weights. But if you try to push the sparsity to 50% or 70% to significantly reduce the model size, performance collapses. ...

Introduction In the rapidly evolving world of Generative AI, we have become accustomed to a specific direction of flow: Text-to-Image (T2I). You type “a futuristic city made of crystal,” and a diffusion model like Stable Diffusion paints it for you. These models are incredibly powerful, having ingested massive datasets that effectively encode a vast amount of “world knowledge.” They know what a city looks like, they know what crystal looks like, and they know how to combine them. ...

Introduction In the rapidly evolving world of Generative AI, one desire stands out above almost all others: Personalization. We all want to put ourselves, our friends, or specific characters into new, imagined worlds. Whether it’s seeing yourself as an astronaut, a cyberpunk warrior, or an oil painting, the goal is high fidelity (it looks exactly like you) and high editability (you can change the background, lighting, and style). For a long time, we have been stuck between two extremes to achieve this: ...

Introduction We are currently witnessing a “Cambrian Explosion” in the world of 3D content generation. With the advent of text-to-image and image-to-3D models, creating a detailed 3D humanoid character used to take an artist days; now, it takes seconds. But there is a massive bottleneck that sits between a static 3D model and a playable video game character: Rigging. Rigging is the digital equivalent of putting a skeleton inside a statue. It involves defining bones (skeleton construction) and telling the computer which parts of the “skin” (the mesh) should move with which bone (skinning). Without rigging, a 3D model is just a statue—it cannot walk, wave, or dance. ...

Introduction Imagine you are driving down a highway. Your eyes are constantly scanning the environment, tracking the speed of the car in front of you, the trees rushing past in your peripheral vision, and the slight drift of your own vehicle. You are performing a complex calculation known in computer vision as Optical Flow estimation—determining how pixels move from one moment to the next. For decades, computer vision researchers have been training AI models to master this task. They use “Ground Truth” data—mathematically perfect calculations of where every pixel actually moved. And modern AI is incredible at this; in many cases, it is far more precise than the human eye. ...

Imagine teaching someone how to repair a bicycle. You rarely give them a list of coordinate geometries or vector rotations. Instead, you show them. You demonstrate how the hand should grip the wrench, the specific twisting motion required, and exactly where the fingers need to apply pressure. In the world of robotics and computer vision, this natural form of instruction—demonstrating the “how”—is incredibly difficult to replicate. Most current systems rely on precise 3D models of objects to plan interactions. But what happens when we want an agent to interact with everyday objects—things that are thin, transparent, deformable, or simply don’t have a pre-existing 3D scan? ...

Introduction In the world of 3D computer vision and graphics, representing shapes accurately is half the battle. While point clouds and meshes are classic formats, Implicit Neural Representations have taken the field by storm. Specifically, Neural Signed Distance Functions (SDFs) have become the gold standard for representing watertight, high-fidelity surfaces. An SDF is a mathematical function that tells you, for any point in 3D space, how far you are from the surface of an object. If you are inside the object, the value is negative; if you are outside, it is positive; and if you are exactly on the surface, the value is zero. ...