Developed and integrated a SAM-based image segmentation feature for the ECLAIR-Robotics/crackle repository, enabling precise extraction of object points within bounding boxes to support advanced robotics perception tasks. Leveraged Python and ROS to implement segment_img.py and connect it with yolo_segment_node.py, facilitating seamless collaboration between segmentation and detection modules. The approach emphasized robustness and reusability, laying groundwork for future SAM-driven enhancements. By automating object segmentation, the work improved perception accuracy and reduced manual labeling requirements, directly benefiting downstream manipulation tasks. The contribution demonstrated practical application of computer vision and image segmentation skills within a robotics context, focusing on measurable workflow improvements.