OmniCam generates camera trajectories from a panorama and text
Original title: OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning
On OmniCaT, reported trajectory error falls 28–47% and collisions 65.8% versus retrained GenDoP.
OmniCam is an autoregressive model that generates camera pose sequences from one panorama and a textual trajectory description. It uses a panoramic point-cloud encoder, hybrid pose tokenization and separate geometric and semantic conditioning with a 3D target anchor. The authors also build OmniCaT, with 267,700 trajectories across four camera behaviors, and test video generation and robotic active perception.
- Against the best baseline, ATE falls 43.0% and collisions 62.3%.
- OmniCaT contains 267,700 trajectories.
- Ablations support geometric and target-aware conditioning.
From the original article
1 min read at huggingface.co
Camera trajectories control viewpoint changes in video generation, scene reconstruction, and robotic perception. Generating them from language requires both scene geometry and target-aware framing. We introduce OmniCam, an autoregressive model that generates camera pose sequences from a single panorama and textual trajectory descriptions.
By Zhenyang Liu, Chenjie Cao, Yisu Zhang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.