A ROS-based Voice Controlled Robotic Arm for Automatic Segregation of Medical Waste Using YOLOv3

The outbreak of the Covid-19 pandemic has resulted in a surge in the generation of medical waste. Due to the transmissible nature of the Virus and the lack of effort at proper disposal, the safety of the front-line health workers, as well as the disposer, is at risk. Hence, to mitigate the spread of infectious diseases, a system is proposed that uses a robotic arm for segregating medical waste automatically. The robotic arm is operable through voice commands, and the segregating operation could function in automatic and manual mode. The system uses the YOLOv3 (You Only Look Once) algorithm to detect and classify the medical waste and then uses the Robot Operating System (ROS) platform to pick up and drop the waste object into color-coded bins. For this research, the medical waste has been categorized into 4 types, and for each type, a color-coded bin has been used for segregation. Our system has achieved 94% training accuracy for the YOLOv3 model on a custom dataset, whereas the system's overall accuracy in automated mode was 82.1%, derived after 30 trials. In the case of manual mode, an average accuracy of 82.5% has been achieved for the same number of trials.