From the model card
Fields below are copied from the tags and counters on the HuggingFace repository Intel/dpt-hybrid-midas at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Intel
- Pipeline tag
- depth-estimation
- Library
- Transformers
- Framework tags
- PyTorch
- License tag
apache-2.0— read the license file in the repo before relying on it- Papers cited
- arXiv:2103.13413
- Downloads (HF counter at last fetch)
- 698,213
- Likes (HF counter at last fetch)
- 110
- Model card
- https://huggingface.co/Intel/dpt-hybrid-midas
Use cases
- Relative depth map generation from photographs
- Depth-guided image editing and compositing
- Scene understanding preprocessing for downstream computer vision tasks
- Robotics or AR applications requiring rough depth estimates
Pros
- Apache-2.0 license
- Transformers depth-estimation pipeline compatible — easy integration
- DPT hybrid architecture balances accuracy and speed
- Published model-index evaluation results
Cons
- Produces relative (not metric) depth — scale information is lost
- DPT-hybrid trails larger MiDaS v3 and Depth Anything models on benchmarks
- Inference speed is slower than lightweight depth models like fast-depth
- Accuracy degrades on textureless surfaces and specular reflections
Tags
transformerspytorchdptdepth-estimationvisionarxiv:2103.13413license:apache-2.0model-indexendpoints_compatibleregion:us