Tutorials

In the following, some tutorials using LightlyTrain can be found. The tutorials are a good starting point to learn about LightlyTrain.

• Classification with Torchvision’s ResNet
• Object Detection with Ultralytics’ YOLOv26
• Monocular Depth Estimation with fastai U-Net (Advanced)
• Embedding Model for Satellite Images with Torchvision’s ResNet
• Debugging Training Instability

Interactive Notebooks

The following Jupyter notebooks demonstrate how to use LightlyTrain for various computer vision tasks. Each notebook can be viewed directly in the documentation or opened in Google Colab for interactive execution.

Core Tasks

• Object Detection with LTDETRv2
• Image Classification with DINOv3
• Depth Estimation with Depth Anything V3
• Semantic Segmentation with DINOv3 EoMT
• Instance Segmentation with DINOv3 EoMT
• Instance Segmentation with LTDETRv2
• Panoptic Segmentation with DINOv3 EoMT

ONNX & TENSORRT EXPORT TUTORIALS

• Object Detection Model Export - LTDETRv2
• Instance Segmentation Model Export - EoMT
• Instance Segmentation Model Export - LTDETRv2
• Semantic Segmentation Model Export - EoMT
• Panoptic Segmentation Model Export - EoMT
• Depth Estimation Model Export (ONNX and TensorRT)

Package Integrations

• LightlyTrain with Ultralytics YOLO
• LightlyTrain with YOLOv12
• LightlyTrain with RF-DETR
• Embedding Model for Satellite Images with Torchvision’s ResNet

Advanced

• Quick Start Distillation Tutorial
• Visualizing ViT Patch Features with PCA