#4848th of 53 in Training and fine-tuning
autodistill
Label images with foundation models, then train small YOLO or DETR detectors
- Stars
- 2.8k
- License
- Apache-2.0
- Last commit
- Sep 2026
- Last release
- Feb 2024
- Language
- Python
2,652 on 21 Apr 20262,799 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Autodistill is a Python package that uses a large base model (Grounded SAM, Grounding DINO, OWLv2 and others) to auto-label unlabeled images from text prompts, then trains a smaller supervised model such as YOLOv8, YOLO-NAS, YOLOv5 or DETR on the result. Base and target models are separate pip plugins, and there is a CLI one-liner as well as a Python API. Supported tasks are object detection and instance segmentation; classification is listed as upcoming.
Who it is for: Computer vision engineers who want custom detectors without hand-labeling data
Strengths
- Modular plugins per base and target model keep dependencies and licenses separate
- Goes from unlabeled images to a trained detector without manual labeling
- One CLI command covers labeling and training: autodistill images --base --target
- Many base models: Grounded SAM 2, OWLv2, Kosmos-2, LLaVA-1.5, cloud vision APIs
Weaknesses
- Vision only; language and other model types are described as future work
- Last release was 2024-02-13, despite recent commits
- Classification and several targets (YOLOv6, YOLOv7) are unfinished or in progress
- Label quality depends on prompts; README says base models are not perfect yet
What it needs
- Needs Python, pip plugins per base/target model, Roboflow (optional, for deployment)
- Models: Grounded SAM, Grounding DINO, Grounded SAM 2, OWLv2, OWL-ViT
Quick answers
Is autodistill open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Also in Training and fine-tuning
See all 53| Rank | Project | Score |
|---|---|---|
| 1 | unslothDesktop and web app to run and fine-tune local models | 88 out of 100 |
| 2 | peftPython library for LoRA and other parameter-efficient fine-tuning of Transformers models | 88 out of 100 |
| 3 | deepspeedPyTorch library for distributed training of very large models with ZeRO | 86 out of 100 |
| #4 | verlReinforcement learning training library for LLM post-training with PPO and GRPO | 86 out of 100 |
| #5 | trlPost-training library with SFT, DPO, GRPO and KTO trainers | 86 out of 100 |