Deep smoke segmentation

Abstract

Inspired by the recent success of fully convolutional networks (FCN) in semantic segmentation, we propose a deep smoke segmentation network to infer high quality segmentation masks from blurry smoke images. To overcome large variations in texture, color and shape of smoke appearance, we divide the proposed network into coarse and fine paths. The coarse path is an encoder-decoder FCN with skip structures, which extracts global context information of smoke and accordingly generates a coarse segmentation mask. To retain fine spatial details of smoke, the fine path is also designed as an encoder-decoder FCN with skip structures, but it is shallower than the coarse path network. Finally, we propose a very small network containing only addition, convolution and activation layers to fuse the results of the two paths. Thus, we can easily train the proposed network end to end for simultaneous optimization of network parameters. To avoid the great difficulty in manually labelling fuzzy smoke boundaries, we propose a method to generate synthetic smoke images. According to the results of our deep segmentation method, we can easily and accurately perform smoke detection on videos. Experiments on three synthetic smoke datasets and one realistic smoke dataset show that our method achieves much better performance than state-of-the-art segmentation algorithms. Test results of our method on videos are also appealing.