Cvpr 2026 Virtual Full Stack

[

Disentangle-then-Align: Non-Iterative Hybrid Multimodal Image Registration via Cross-Scale Feature Disentanglement.

This video presents our

[CVPR 2026] Think-Then-Generate: Structural Chain-of-Thought Reasoning for Consistent3D Generation

In bandwidth-limited

TAPE: Task-Adaptive Prototype Evolution in Audio-Language Models for Fully Few-shot Class-incremental Audio Classification.

Leon Liangyu Chen, Haoyu Ma, Zhipeng Fan, Ziqi Huang, Animesh Sinha, Xiaoliang Dai, Jialiang Wang, Zecheng He, Jianwei ...

MixerCSeg: An Efficient Mixer Architecture for Crack Segmentation via Decoupled Mamba Attention.

Paper: Bootstrapping Multi-view Learning for Test-time Noisy Correspondence Authors: Changhao He, Di Xue, Shuxian Li, Yanji ...

VIMCAN: Visual-Inertial 3D Human Pose Estimation with Hybrid Mamba-Cross-Attention Network.

arxiv.org/abs/2512.14884.

NeuroFlow: Toward Unified Visual Encoding and Decoding from Neural Activity.

Hyun Lee, Hyemin Jeong, Yejin Kim, Hyungwook Choi, Hyunsoo Cho, Soo Kyung Kim, Joonseok Lee. A More Word-like Image ...

Paper: https://arxiv.org/abs/2512.01390 Project Page: https://cmlab-korea.github.io/FRAMER/ Authors/Affiliations: [Seungho ...

CVPR 2026: Learning 3D Shape Fidelity Metric from Real-world Distortions

The introduction of the

[CVPR 2026]SFR-Net: Steering-Fusion-Refining Network in Multi-label Zero-Shot Sewer Defect Detection

In this video, we introduce our

Text-guided diffusion models have advanced image editing by enabling intuitive control through language. However, despite their ...