• We have release a webui version using the Segment Anything Model!
  Please check out here https://github.com/derekray311511/SAM-webui

The code requires python>=3.8, as well as pytorch>=1.7 and torchvision>=0.8. Please follow the instructions here to install both PyTorch and TorchVision dependencies. Installing both PyTorch and TorchVision with CUDA support is strongly recommended.

We have tested the setting below on 4090, 3060ti, 1060-6G
Python 3.8 pytorch 2.0.0 (py3.8_cuda11.7_cudnn8.5.0_0) torchvision 0.15.0

Install Segment Anything:

https://github.com/derekray311511/segment-anything.git
cd segment-anything; pip install -e .

The following optional dependencies are necessary for mask post-processing, saving masks in COCO format, the example notebooks, and exporting the model in ONNX format. jupyter is also required to run the example notebooks.

pip install opencv-python pycocotools matplotlib onnxruntime onnx

You can download the model checkpoints here.

python scripts/select_obj.py --img /PATH/TO/YOUR/IMG.file_type --output /OUTPUT/FILE/NAME --model_type MODEL_TYPE --checkpoint /PATH/TO/MODEL

MODEL_TYPE: vit_h, vit_l, vit_b

• Auto: Segment all objects in the image
• Custom: Select object(s) with points or boxes using mouse clicks
• View: View the masks you just created and disable manipulation

• Press SPACE to inference all objects in the image

• Point select: Press p to switch to point select function
  • a: Positive prompt
  • d: Negative prompt
• Box select: Press b to switch to box select function

• Press v to switch between view / previous mode