Shuai Zhang, Zhanpeng Ouyang, Zhengqing Chen, Ruohong Mei, Shaoyu Chen, Weiqiang Ren, Hui Sun, Qian Zhang · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28521 · open-access PDF
Generasi video dan rekonstruksi feed-forward 3D biasanya menggunakan representasi terpisah, membutuhkan dekoding RGB dan re-encoding visual untuk memulihkan geometri dari latent yang dihasilkan. Kami mengusulkan Latent Bridge, yang menghubungkan ruang latent yang dihasilkan oleh video dengan ruang fitur geometris untuk secara langsung memprediksi geometri 3D dari latent video. Kami mengembangkan jembatan spesifik representasi untuk Wan2.1 dan V-RAE, memungkinkan prediksi feed-forward kamera, kedalaman, dan awan titik 3D tanpa jalur RGB menengah.
Machine translation (Meta M2M100); the English original follows.
Video generation and feed-forward 3D reconstruction typically use separate representations, requiring RGB decoding and visual re-encoding to recover geometry from generated latents. We propose Latent Bridge, which connects video generative latent spaces with geometric feature spaces to directly predict 3D geometry from video latents. We develop representation-specific bridges for Wan2.1 and V-RAE, enabling feed-forward prediction of cameras, depth, and 3D point clouds without the intermediate RGB pathway. Each bridge is trained on RGB videos through feature distillation, while the video encode
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28521, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange