Emotion being a subjective thing, leveraging knowledge and science behind labeled data and extracting the components that constitute it. With the development of deep learning in computer vision, emotion recognition has become a widely-tackled research problem. For mobility and privacy reasons, the required image processing should be local on embedded computer platforms with performance requirements and energy constraints. For this purpose, in this work, we propose a Field Programmable Gate Array (FPGA) architecture applied for this task using independent method called convolutional neural network (CNN [1]). The design flow is evaluated by implementing the previously trained CNN to recognize facial emotions from face image implemented in python on a PC. The project explains the process of porting the CNN algorithm from python to C/C++ and then executing it on a ZYNQ FPGA board. Once we have trained a network, weights from the Tensorflow model will be convert as C-arrays. After having the weights as C arrays, they can be implemented to FPGA system. This method was trained on the posed-emotion dataset (FER2013). The results show that with more fine-tuning and depth, the CNN model can outperform the state-of-the-art methods for emotion recognition. The bottleneck of the CNN [2] is the convolutional layers and that is why different solutions for that accelerator are analyzed and the performance of each solution is tested.
CITATION STYLE
Le-Tien, T., Phan-Xuan, H., & Nguyen-Tan, S. (2019). Light-Weight Deep Convolutional Network-Based Approach for Recognizing Emotion on FPGA Platform. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11814 LNCS, pp. 287–306). Springer. https://doi.org/10.1007/978-3-030-35653-8_19
Mendeley helps you to discover research relevant for your work.