#4242nd of 70 in Open modelsFor image, video and design, 3d
vipe
Estimates camera pose, intrinsics and dense depth from raw video
- Stars
- 2.2k
- License
- custom license
- Last commit
- Nov 2026
- Last release
- Jun 2026
- Language
- Python
1,845 on 21 Apr 20262,200 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
ViPE is a Python pipeline from NVIDIA that estimates camera intrinsics, camera motion and near-metric dense depth from unconstrained video, including wide-angle and 360-degree panorama footage. It installs from PyPI (nvidia-vipe) and runs through a `vipe infer` CLI. Newer pipelines handle long videos with bounded memory (`pose_only_long`) and streaming metric reconstruction with loop closure and point cloud export (`metrictrail`).
Who it is for: Researchers annotating video with camera poses and depth for 3D or robotics data
Strengths
- Handles pinhole, wide-angle and 360-degree panorama video in one tool
- `pose_only_long` keeps GPU and CPU memory roughly constant regardless of video length
- Outputs pose, intrinsics and depth per video; MetricTrail adds PLY point cloud export
- Release 1.2.0 reports a 2.7x speed-up from CUDA fused kernels and caching
Weaknesses
- Mixed licensing: Apache 2.0 source, but Unik3D is BY-NC-SA 4.0 and MetricTrail weights are noncommercial
- MetricTrail needs custom CUDA kernel builds, a pinned xformers and CUDA 12.8 toolchain
- MetricTrail weights are obtained separately; the Hugging Face release is described as planned
- Loop closure needs external SALAD candidates exported from a separate environment
What it needs
- GPU required
- Needs CUDA 12.8, conda, uv, xformers 0.0.33.post1, PyPose (optional, loop closure), SALAD (optional, loop closure)
- Models: MoGe v2 (moge2-l), Depth-Anything 3, DINO-DPVO, DINOv3 ViT-S/16, Unik3D
Quick answers
Is vipe open source?
Its code is public on GitHub under its own license terms; read them before you use it at work.
Does vipe need a GPU?
Yes, it needs a GPU.
Also in Open models
See all 70| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 79 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 76 out of 100 |