Developed and integrated an automated trash detection feature for the RoBorregos/home2 repository, focusing on enhancing the vision pipeline to support waste-management workflows. Leveraging Python, ROS, and computer vision techniques, the work centered on implementing a YOLO-E-based detection node that identifies and classifies trash in images, seamlessly connecting with existing vision tasks. The project included refactoring service interfaces, normalizing bounding box data, and improving output accuracy to ensure reliable detection results. Emphasis was placed on data integrity, developer experience, and cross-team collaboration, with updates to code style and alignment across related tasks to maintain consistency and facilitate ongoing development.