AI Frontiers Dec 30, 2025: 100 Papers on 3D Scene Sculpting & Video Rewinding скачать в хорошем качестве

AI Frontiers Dec 30, 2025: 100 Papers on 3D Scene Sculpting & Video Rewinding

This video synthesizes 100 computer vision research papers from arXiv on December 30, 2025, revealing a major shift from passive perception to active creation. Key themes include: 1) *Generative & Controllable Scene Synthesis* - Tools like SpaceTimePilot enable re-rendering dynamic scenes from new viewpoints with altered motion, treating video as a navigable 4D fabric. 2) *3D Scene Understanding & Reconstruction* - Methods like GaMO achieve state-of-the-art 3D reconstruction with 25x speedups by reformulating the problem as multi-view outpainting. 3) *Instant 3D Editing* - Edit3r allows language-driven 3D scene edits from sparse, unposed photos in seconds, breaking the optimization bottleneck. 4) *Robustness in Real Conditions* - Research addresses low-light vision (DarkEQA) and corrupted data (FineTec) for real-world deployment. 5) *Multimodal Integration* - Language becomes the control interface for vision, with benchmarks evaluating the reasoning behind video generation (VIPER). Dominant methodologies include diffusion models for generation, 3D Gaussian splatting for efficient rendering, transformer architectures for multimodal integration, and parameter-efficient fine-tuning (LoRA) for model specialization. The collective research points toward future general-purpose spatial foundation models that combine perception, reconstruction, and generation. **AI Synthesis Process**: This content was created using AI tools including GPT models (specifically deepseek-chat) for analysis and synthesis of the research summaries, text-to-speech synthesis using Amazon Polly for audio narration, and image generation using Google's Imagen for visual illustrations. The process involved extracting key insights from 100 paper abstracts, identifying overarching themes, and presenting them in an accessible narrative format. 1. Zhening Huang et al. (2025). SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time. https://arxiv.org/pdf/2512.25075v1 2. Yi-Chuan Huang et al. (2025). GaMO: Geometry-aware Multi-view Diffusion Outpainting for Sparse-View 3D Reconstruction. https://arxiv.org/pdf/2512.25073v1 3. Jiageng Liu et al. (2025). Edit3r: Instant 3D Scene Editing from Sparse Unposed Images. https://arxiv.org/pdf/2512.25071v1 4. Dian Shao et al. (2025). FineTec: Fine-Grained Action Recognition Under Temporal Corruption via Skeleton Decomposition and Sequence Completion. https://arxiv.org/pdf/2512.25067v1 5. Xu He et al. (2025). From Inpainting to Editing: A Self-Bootstrapping Framework for Context-Rich Visual Dubbing. https://arxiv.org/pdf/2512.25066v1 6. Yuchen Wu et al. (2025). FoundationSLAM: Unleashing the Power of Depth Foundation Models for End-to-End Dense Visual SLAM. https://arxiv.org/pdf/2512.25008v1 7. Zhenyu Cui et al. (2025). Bi-C2R: Bidirectional Continual Compatible Representation for Re-indexing Free Lifelong Person Re-identification. https://arxiv.org/pdf/2512.25000v1 8. Yohan Park et al. (2025). DarkEQA: Benchmarking Vision-Language Models for Embodied Question Answering in Low-Light Indoor Environments. https://arxiv.org/pdf/2512.24985v1 9. Itallo Patrick Castro Alves Da Silva et al. (2025). Evaluating the Impact of Compression Techniques on the Robustness of CNNs under Natural Corruptions. https://arxiv.org/pdf/2512.24971v1 10. Siyuan Hu et al. (2025). ShowUI-$π$: Flow-based Generative Models as GUI Dexterous Hands. https://arxiv.org/pdf/2512.24965v1 11. Yifan Li et al. (2025). VIPER: Process-aware Evaluation for Generative Video Reasoning. https://arxiv.org/pdf/2512.24952v1 12. Xinran Gong et al. (2025). ProDM: Synthetic Reality-driven Property-aware Progressive Diffusion Model for Coronary Calcium Motion Correction in Non-gated Chest CT. https://arxiv.org/pdf/2512.24948v1 13. Wentao Zhang et al. (2025). CPJ: Explainable Agricultural Pest Diagnosis via Caption-Prompt-Judge with LLM-Judged Refinement. https://arxiv.org/pdf/2512.24947v1 14. Rongji Xun et al. (2025). HaineiFRDM: Explore Diffusion to Restore Defects in Fast-Movement Films. https://arxiv.org/pdf/2512.24946v1 15. Bartłomiej Olber et al. (2025). Semi-Supervised Diversity-Aware Domain Adaptation for 3D Object detection. https://arxiv.org/pdf/2512.24922v1 16. Zichen Tang et al. (2025). FinMMDocR: Benchmarking Financial Multimodal Reasoning with Scenario Awareness, Document Understanding, and Multi-Step Computation. https://arxiv.org/pdf/2512.24903v1 17. Meng Lan et al. (2025). OFL-SAM2: Prompt SAM2 with Online Few-shot Learner for Efficient Medical Image Segmentation. https://arxiv.org/pdf/2512.24861v1 18. Xunyi Zhao et al. (2025). VLN-MME: Diagnosing MLLMs as Language-guided Visual Navigation agents. https://arxiv.org/pdf/2512.24851v1 Disclaimer: This video uses arXiv.org content under its API Terms of Use; AI Frontiers is not affiliated with or endorsed by arXiv.org.