SAMA Light Annotator

Labeling Images for Object Detection, Instance Segmentation and Semantic Segmentation

Install

For light version without SAM run:

1. pip install -r requirements_light.txt
2. python annotator_light.py

For version with SAM run:

1. pip install -r requirements_sam.txt
2. python annotator_sam.py

SAM settings

1. Set SAM_WEIGHT_PATH, SAM_MODEL and SAM_PLATFORM in utils.ml_config.py. By default SAMA Light use SAM2_large on 'cuda'
2. List of supported methods: SAM, SAM-HQ, SAM-2, MobileSAM, EfficientSAM, FastSAM
3. For models SAM2, FastSAM weights will be downloaded automatically.
4. For other models you have to download weight manually and place it according ml_config.py SAM_WEIGHT_PATH variable. You can download weights from:

• SAM_HQ
• FastSAM
• SAM_2
• SAM
• MobileSAM
• EfficientSAM