A Tensorflow Based Robotic Arm

Object detection capability is provided to the Robotic arm using the Tensorflow object detection API. Tensorflow is having efficient deep learning libraries for object detection. This paper mainly deals with object detection in a real-time and provided the visual data to the robotic arm to choose the object. Object detection training is done under a supervised learning method by providing label images. Faster RCNN inception v2 model is used for the customized training of the toy images. Voice input through a microphone is taken and processed to convert it into a text to get object name needed to pick by the arm. Real-time video or image is captured by using the web camera or mobile IP camera application, sends to the system to process and detect the object in a real time. If the object found in the image, it calculates the physical distance and sends the data to the Arduino board connected with the robotic arm in term of the degree of rotation to access the object.