# LightlyTrain Documentation
```{eval-rst}
.. image:: _static/lightly_train_light.svg
:align: center
:class: only-light
.. image:: _static/lightly_train_dark.svg
:align: center
:class: only-dark
```
[](https://colab.research.google.com/github/lightly-ai/lightly-train/blob/main/examples/notebooks/object_detection.ipynb)
[](https://docs.lightly.ai/train/stable/installation.html)
[](https://docs.lightly.ai/train/stable/installation.html)
[](https://docs.lightly.ai/train/stable/docker.html#)
[](https://docs.lightly.ai/train/stable/)
[](https://discord.gg/xvNJW94)
*Train Better Models, Faster*
LightlyTrain is the leading framework for transforming your data into state-of-the-art
computer vision models. It covers the entire model development lifecycle from
pretraining DINOv2/v3 vision foundation models on your unlabeled data to fine-tuning
transformer and YOLO models on detection and segmentation tasks for edge deployment.
[Contact us](https://www.lightly.ai/contact) to request a license for commercial use.
Also check out [LightlyStudio](https://github.com/lightly-ai/lightly-studio) to easily
visualize your annotations and predictions.
## News
- \[[0.15.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-15-0)\] -
2026-04-14: 🔎 **Distillationv3:** Better generalizing distillation method that
performs equally well across dense and global tasks and across all models, from ViTs
to hybrids to CNNs (+support for custom teachers!). 🔎
- \[[0.14.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-14-0)\] -
2026-01-19: 🐣 **PicoDet, Tiny Models, and ONNX/TensorRT FP16 Support:** PicoDet object
detection models for low-power embedded devices! All tasks now support tiny DINOv3
models and ONNX/TensorRT export in FP16 precision for faster inference! 🐣
- \[[0.13.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-13-0)\] -
2025-12-15: 🐥 **New Tiny Object Detection Models:** We release tiny DINOv3 models
pretrained on COCO for
[object detection](https://docs.lightly.ai/train/stable/object_detection.html#coco)! 🐥
- \[[0.12.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-12-0)\] -
2025-11-06: 💡 **New DINOv3 Object Detection:** Run inference or fine-tune DINOv3
models for
[object detection](https://docs.lightly.ai/train/stable/object_detection.html)! 💡
- \[[0.11.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-11-0)\] -
2025-08-15: 🚀 **New DINOv3 Support:** Pretrain your own model with
[distillation](https://docs.lightly.ai/train/stable/pretrain_distill/methods/distillation.html#methods-distillation)
from DINOv3 weights. Or fine-tune our SOTA
[EoMT semantic segmentation model](https://docs.lightly.ai/train/stable/semantic_segmentation.html#semantic-segmentation-eomt-dinov3)
with a DINOv3 backbone! 🚀
- \[[0.10.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-10-0)\] -
2025-08-04: 🔥 **Train state-of-the-art semantic segmentation models** with our new
[**DINOv2 semantic segmentation**](https://docs.lightly.ai/train/stable/semantic_segmentation.html)
fine-tuning method! 🔥
- \[[0.9.0](https://docs.lightly.ai/train/stable/changelog.html#changelog-0-9-0)\] -
2025-07-21:
[**DINOv2 pretraining**](https://docs.lightly.ai/train/stable/pretrain_distill/methods/dinov2.html)
is now officially available!
## Workflows
````{grid} 1 1 2 3
```{grid-item-card} Object Detection
:link: object_detection.html
Train LTDETR detection models with DINOv2, DINOv3, or EdgeCrafter ECViT backbones.
```
```{grid-item-card} Instance Segmentation
:link: instance_segmentation.html
Train EoMT segmentation models with DINOv3 backbones.
```
```{grid-item-card} Panoptic Segmentation
:link: panoptic_segmentation.html
Train EoMT segmentation models with DINOv2 or DINOv3 backbones.
```
```{grid-item-card} Semantic Segmentation
:link: semantic_segmentation.html
Train EoMT segmentation models with DINOv2 or DINOv3 backbones.
```
```{grid-item-card} Image Classification
:link: image_classification.html
Train image classification models with any backbone.
```
```{grid-item-card} Depth Estimation (NEW)
:link: depth_estimation.html
Run Depth Anything V2 and V3 monocular depth inference.
```
```{grid-item-card} Distillation
:link: pretrain_distill/methods/distillation.html
Distill knowledge from DINOv2 or DINOv3 into any model architecture.
```
```{grid-item-card} Pretraining
:link: pretrain_distill/methods/dinov2.html
Pretrain DINOv2 foundation models on your domain data.
```
```{grid-item-card} Autolabeling
:link: predict_autolabel.html
Generate high-quality pseudo labels for detection and segmentation tasks.
```
````
## How It Works [](https://colab.research.google.com/github/lightly-ai/lightly-train/blob/main/examples/notebooks/object_detection.ipynb)
Install Lightly**Train** on Python 3.8+ for Windows, Linux or MacOS.
```bash
pip install lightly-train
```
Then train an object detection model with:
```python
import lightly_train
if __name__ == "__main__":
lightly_train.train_object_detection(
out="out/my_experiment",
model="ltdetrv2-s-coco",
data={
# ... Data configuration
}
)
```
And run inference like this:
```python
import lightly_train
# Load the model from the best checkpoint
model = lightly_train.load_model("out/my_experiment/exported_models/exported_best.pt")
# Or load one of the models hosted by LightlyTrain
model = lightly_train.load_model("ltdetrv2-s-coco")
results = model.predict("image.jpg")
```
See the full [quick start guide](quick-start-object-detection) for more details.
## Features
- Python, Command Line, and [Docker](https://docs.lightly.ai/train/stable/docker.html)
support
- Built for
[high performance](https://docs.lightly.ai/train/stable/performance/index.html)
including [multi-GPU](https://docs.lightly.ai/train/stable/performance/multi_gpu.html)
and [multi-node](https://docs.lightly.ai/train/stable/performance/multi_node.html)
support
- [Monitor training progress](https://docs.lightly.ai/train/stable/pretrain_distill.html#logging)
with MLflow, TensorBoard, Weights & Biases, and more
- Runs fully on-premises with no API authentication
- Export models in their native format for fine-tuning or inference
- Export models in ONNX or TensorRT format for edge deployment
## Models
LightlyTrain supports the following model and workflow combinations.
### Fine-tuning
| Model | Object
Detection | Instance
Segmentation | Panoptic
Segmentation | Semantic
Segmentation | Image
Classification |
| ----------- | :-----------------------------------------------------------------: | :----------------------------------------------------------------------: | :----------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------: |
| DINOv3 | ✅ [🔗](https://docs.lightly.ai/train/stable/object_detection.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/instance_segmentation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/panoptic_segmentation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/semantic_segmentation.html#use-eomt-with-dinov3) | ✅ [🔗](https://docs.lightly.ai/train/stable/image_classification.html) |
| DINOv2 | ✅ [🔗](https://docs.lightly.ai/train/stable/object_detection.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/instance_segmentation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/panoptic_segmentation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/semantic_segmentation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/image_classification.html) |
| EdgeCrafter | ✅ [🔗](https://docs.lightly.ai/train/stable/object_detection.html) | | | | |
| Any | | | | | ✅ [🔗](https://docs.lightly.ai/train/stable/image_classification.html) |
### Distillation & Pretraining
| Model | Distillation | Pretraining |
| ------------------------------------------ | :----------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------: |
| DINOv3 | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/methods/distillation.html#distill-from-dinov3) | |
| DINOv2 | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/methods/distillation.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/methods/dinov2.html) |
| Torchvision ResNet, ConvNext, ShuffleNetV2 | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/torchvision.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/torchvision.html) |
| TIMM models | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/timm.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/timm.html) |
| Ultralytics YOLOv5–YOLO12, RT-DETR | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/ultralytics.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/ultralytics.html) |
| RT-DETR, RT-DETRv2 | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/rtdetr.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/rtdetr.html) |
| RF-DETR | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/rfdetr.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/rfdetr.html) |
| YOLOv12 | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/yolov12.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/yolov12.html) |
| Custom PyTorch Model | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/custom_models.html) | ✅ [🔗](https://docs.lightly.ai/train/stable/pretrain_distill/models/custom_models.html) |
[Contact us](https://www.lightly.ai/contact) if you need support for additional models.
## Usage Events
LightlyTrain collects anonymous usage events to help us improve the product. We only
track training method, model architecture, and system information (OS, GPU, CI,
Container). To opt-out, set the environment variable:
`export LIGHTLY_TRAIN_EVENTS_DISABLED=1`
## License
Lightly**Train** offers flexible licensing options to suit your specific needs:
- **AGPL-3.0 License**: Perfect for open-source projects, academic research, and
community contributions. Share your innovations with the world while benefiting from
community improvements.
- **Commercial License**: Ideal for businesses and organizations that need proprietary
development freedom. Enjoy all the benefits of LightlyTrain while keeping your code
and models private. Includes model training and runtime license.
- **Free Community License**: Available for students, researchers, startups in early
stages, or anyone exploring or experimenting with LightlyTrain. Empower the next
generation of innovators with full access to the world of pretraining.
### Commercial Pricing
| Plan | Price | Eligibility |
| -------------- | --------------- | -------------------------------------- |
| **Startup** | \$5,000 / year | < \$1M revenue **or** < 10 employees |
| **Growth** | \$10,000 / year | < \$10M revenue **or** < 100 employees |
| **Enterprise** | Custom | > \$10M revenue **or** > 100 employees |
All commercial plans include a license for model training, edge deployment, and
inference. See the full [Licensing](licensing.md) page for plan details and inclusions.
[Contact us](https://www.lightly.ai/contact) to get started — we'll find the right
option for your project!
## Contact
[](https://www.lightly.ai/lightly-train)
[](https://discord.gg/xvNJW94)
[](https://github.com/lightly-ai/lightly-train)
[](https://x.com/lightlyai)
[](https://www.linkedin.com/company/lightly-tech)
```{toctree}
---
hidden:
maxdepth: 2
---
quick_start_object_detection
quick_start_distillation
installation
object_detection
instance_segmentation
panoptic_segmentation
semantic_segmentation
image_classification
depth_estimation
pretrain_distill/index
predict_autolabel
embed
Settings
data/index
performance/index
docker
tutorials/index
python_api/index
faq
licensing
changelog
```
