isl-org/DPT

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

2334 stars · 281 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DPT (Dense Prediction Transformers) is Intel's implementation of Vision Transformers for dense prediction tasks, providing pretrained models for monocular depth estimation and semantic segmentation. It is Python/PyTorch-based inference code released alongside the paper 'Vision Transformers for Dense Prediction'.

Use cases

• estimate depth from a single image
• run monocular depth estimation on photos
• perform semantic segmentation with a vision transformer
• get pretrained DPT models for depth and segmentation
• use vision transformers for dense prediction tasks
• depth estimation for autonomous driving datasets like KITTI

When to choose

• you need pretrained transformer-based monocular depth estimation models
• you want semantic segmentation with ViT-based architectures
• you need a reference implementation of the DPT paper
• you want depth models finetuned on KITTI or NYUv2

When to avoid

• you need actively maintained software with bug fixes and updates
• you require production support or new model releases
• you want a project that accepts patches or contributions
• you need compatibility with recent PyTorch versions without adaptation

Facets

library · maturity abandoned

machine-learning deep-learning computer-vision image-processing computer-vision deep-learning machine-learning artificial-intelligence python cross-platform vision-transformer depth-estimation semantic-segmentation monodepth pytorch dense-prediction discontinued

Member repositories

For agents

markdown · JSON · MCP: product_card(name="isl-org/DPT")