The project aimed to develop a semi-autonomous pipeline to generate grasp poses for objects scattered in the environment visible to robots. Once objects in the environment are identified, the operator chooses which object to grasp. After the operator chooses the desired object, the pipeline registers the object's point cloud, segments the point cloud into its parts, and generates a grasp pose for the particular object and then goes to grasp the desired object.

Object Identification from YOLO-v7

Object Pointcloud Generation

Segmented Cup Point Cloud

Segmented Hammer Point Cloud

Generated grasp poses for each part

Generated grasp poses for each part