The model was created on a computer using the Google TensorFlow frame- work and the Keras library. The activation function "ReLU" and the optimizer "Adam" in the dense layers have a learning rate of 3e-4 each. Since this is a binary classification problem, binary cross-entropy was employed...
Data analysis was carried out on a workstation to reduce the computational load of the social robot. At the same time, both the MAT social robot and the workstation were installed with OpenCV 3.1, TensorFlow 0.9, the proposed F-YOLO, and B-RPN. environment. The MAT social robot processing...