Publication

Abstract : In today’s fast paced logistics and manufacturing sectors, being efficient and accurate is crucial. An industrial and warehouse-oriented computer vision-enabled pick-and-place robot is the subject of this research. To accurately identify, grab, and transfer goods inside a work area, the proposed system merges advanced computer vision technology with robotic arms. The robot can detect and operate objects of varied sizes, shapes, and positions owing to machine learning. This ability of the robot improves efficiency and accuracy while eliminating the need for humans to do repetitive tasks. Training the computer vision model on a custom dataset, assembling the robotic system, and testing the system in real-world circumstances are all part of the work. This work aims to develop a simple yet efficient pick and place robotic system. A Raspberry Pi is used as the central processing unit for the robot. The object detection model deployed on the microcontroller is trained on an extensively designed dataset, with various augmentations. In order to choose an optimal object detection model for the work, a model performance evaluation was conducted based on different training metrics like accuracy and processing speed on different devices. A comparative study of Yolov5 and SSD models is also performed and the accuracy, and efficiency of the algorithms were analyzed.