Deep convolutional Neural Networks (CNN) are the state-of-the-art performersfor object detection task. It is well known that object detection requires morecomputation and memory than image classification. Thus the consolidation of aCNN-based object detection for an embedded system is more challenging. In thiswork, we propose LCDet, a fully-convolutional neural network for generic objectdetection that aims to work in embedded systems. We design and develop anend-to-end TensorFlow(TF)-based model. Additionally, we employ 8-bitquantization on the learned weights. We use face detection as a use case. OurTF-Slim based network can predict different faces of different shapes and sizesin a single forward pass. Our experimental results show that the proposedmethod achieves comparative accuracy comparing with state-of-the-art CNN-basedface detection methods, while reducing the model size by 3x and memory-BW by~4x comparing with one of the best real-time CNN-based object detector such asYOLO. TF 8-bit quantized model provides additional 4x memory reduction whilekeeping the accuracy as good as the floating point model. The proposed modelthus becomes amenable for embedded implementations.
展开▼
