🏠 Getting Started

To get started, follow the installation steps below. You can install the components locally or use Docker for a streamlined process.

Begin by installing AnyGrasp. The detailed installation instructions can be found here.

Once AnyGrasp is set up, proceed to install SAM. Follow the instructions provided here.

Now you can run the code that uses AnyGrasp SDK & Grounded-Light-HQ-SAM.

cd TargetedAnyGrasp
conda activate sam
cd grasp_detection
python sample_realsense.py
# s - save
# q - exit

Then you can find the image (RGB+depth) captured in the "out" folder named by timestamps.

conda activate sam
cd Grounded-Segment-Anything/EfficientSAM
python grounded_light_hqsam.py

Then you will get the grounded detection and segmentation results under ./grasp_detection/out/.

conda activate sam
cd grasp_detection

# for mask
python select_mask_region.py

# for bounding box
python select_region.py

conda activate anygrasp
cd grasp_detection
sh demo_real_query.sh

Finally, you will get the gripper_pose.json.

🎬 Demo

You can run the code for an example scene and the queried case - "bear" (under ./grasp_detection/out/1/).