Computer Vision

Toto, a new autoregressive video model, achieves competitive performance across various benchmarks by pre-training on over 1 trillion visual tokens, demonstrating the effectiveness of scaling video mo…

SPAR3D: Fast, accurate single-image 3D reconstruction via a novel two-stage approach using point clouds for high-fidelity mesh generation.

This paper unveils critical thresholds for efficient visual autoregressive model computation, proving sub-quartic time is impossible beyond a certain input matrix norm while establishing efficient app…

MoDec-GS: a novel framework achieving 70% model size reduction in dynamic 3D Gaussian splatting while improving visual quality by cleverly decomposing complex motions and optimizing temporal intervals…

DOLPHIN: AI automates scientific research from idea generation to experimental validation.

Diffusion as Shader (DaS) achieves versatile video control by using 3D tracking videos as control signals in a unified video diffusion model, enabling precise manipulation across diverse tasks.

Chirpy3D: Generating creative, high-quality 3D birds with intricate details by learning a continuous part latent space from 2D images.

TransPixar generates high-quality videos with transparency by jointly training RGB and alpha channels, outperforming sequential generation methods.

Through-The-Mask uses mask-based motion trajectories to generate realistic videos from images and text, overcoming limitations of existing methods in handling complex multi-object motion.

STAR: A novel approach uses text-to-video models for realistic, temporally consistent real-world video super-resolution, improving image quality and detail.

MotionBench, a new benchmark, reveals that existing video models struggle with fine-grained motion understanding. To address this, the authors propose TE Fusion, a novel architecture that improves mo…

GS-DiT: Generating high-quality videos with advanced 4D control through efficient dense 3D point tracking and pseudo 4D Gaussian fields.

DepthMaster tames diffusion models for faster, more accurate monocular depth estimation by aligning generative features with high-quality semantic features and adaptively balancing low and high-freque…

MagicFace achieves high-fidelity facial expression editing via AU control, preserving identity and background using a diffusion model and ID encoder, significantly outperforming existing methods.

Ingredients: A new framework customizes videos by blending multiple photos with video diffusion transformers, enabling realistic and personalized video generation while maintaining consistent identity…

VideoAnydoor: High-fidelity video object insertion with precise motion control, achieved via an end-to-end framework leveraging an ID extractor and a pixel warper for robust detail preservation and fi…

SeFAR: a novel semi-supervised framework for fine-grained action recognition, achieves state-of-the-art results by using dual-level temporal modeling, moderate temporal perturbation, and adaptive regu…

SeedVR: A novel diffusion transformer revolutionizes generic video restoration by efficiently handling arbitrary video lengths and resolutions, achieving state-of-the-art performance.

LightningDiT resolves the optimization dilemma in latent diffusion models by aligning latent space with pre-trained vision models, achieving state-of-the-art ImageNet 256x256 generation with over 21x …

Zero-shot image safety judgment is achieved using MLLMs and a novel method called CLUE, objectifying safety rules, and significantly reducing the need for human labeling.