IFQ-Net: Integrated Fixed-Point Quantization Networks for Embedded Vision

Hongxing Gao, Wei Tao, Dongchao Wen, Tse-Wei Chen, Kinya Osa, Masami Kato; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 607-615

Deploying deep models on embedded devices has been a challenging problem since the great success of deep learning based networks. Fixed-point networks, which represent their data with low bits fixed-point and thus give remarkable savings on memory usage, are generally preferred. Even though current fixed-point networks employ relative low bits (e.g. 8-bits), the memory saving is far away from enough for the embedded devices. On the other hand, quantization deep networks, for example XNOR-Net and HWGQ-Net, quantize the data into 1 or 2 bits resulting in more significant memory savings but still contain lots of floating-point data. In this paper, we propose a fixed-point network for embedded vision tasks through converting the floating-point data in a quantization network into fixed-point. Furthermore, to overcome the data loss caused by the conversion, we propose to compose floating-point data operations across multiple layers (e.g. convolution, batch normalization and quantization layers) and convert them into fixed-point. We name the fixed-point network obtained through such integrated conversion as Integrated Fixed-point Quantization Networks (IFQ-Net). We demonstrate that our IFQ-Net gives 2.16x and 18x more savings on model size and runtime feature map memory respectively with similar accuracy on ImageNet. Furthermore, based on YOLOv2, we design IFQ-Tinier-YOLO face detector which is a fixed-point network with 256x reduction in model size (246k Bytes) than Tiny-YOLO. We illustrate the promising performance of our face detector in terms of detection rate on Face Detection Data Set and Benchmark (FDDB) and qualitative results of detecting small faces of Wider Face dataset.