Fully convolutional line parsing

We present a one-stage F ully C onvolutional Li ne P arsing network (F-Clip) that detects line segments from images. The proposed is very simple and flexible with variations gracefully trade off between speed accuracy for different applications. F-Clip in an end-to-end fashion by predicting each line’s center position, length, angle. further customize the design of convolution kernels our fully convolutional to effectively exploit statistical priors distribution angles real image datasets. conduct extensive experiments show method achieves significantly better trade-off efficiency accuracy, resulting real-time detector at up 73 FPS on single GPU. Such inference makes readily applicable tasks without compromising any previous methods. Moreover, when equipped performance-improving backbone network, able outperform all state-of-the-art detectors similar or even higher frame rate. In other word, under same speed, always achieving best compare Source code https://github.com/Delay-Xili/F-Clip .