基于深度学习的图像实例分割
|
摘要
提出了一种基于深度学习的精确图像分割方法。在Mask-R-CNN网络基础上给出一种实例分割网络。针对精确边界分割问题,提出通过重新设计掩码分支结构,来改善边界分割精度。在掩码分支上使用了前后层特征融合的方法可以更好的保留边缘信息。进一步通过增大RoIAlign层的分辨率,得到了更加精确的边界信息。在不影响算法精度的前提下采用深度可分离卷积减少了训练参数,提高了分割算法的效率。实验时通过比较mAP(平均准确率)的结果和检测定量图片需要的时间以及消耗的内存表明该算法的精确性和高效性。
An accurate image segmentation method based on deep learning is proposed.An instance segmentation network is given based on the Mask-R-CNN network.For the problem of precise boundary segmentation,it is proposed to improve the boundary segmentation accuracy by redesigning the mask branch structure.The use of front-and-back layer feature fusion on the mask branch can better preserve the edge information.By increasing the resolution of the RoIAlign layer,more accurate boundary information is obtained.The use of depthwise separable convolution reduces the training parameters without affecting the accuracy of the algorithm,and improves the efficiency of the segmentation algorithm.The experiment demonstrates the accuracy and efficiency of the algorithm by comparing the results of mAP(mean average precision)with the time required to detect quantified pictures and the memory consumed.
An accurate image segmentation method based on deep learning is proposed.An instance segmentation network is given based on the Mask-R-CNN network.For the problem of precise boundary segmentation,it is proposed to improve the boundary segmentation accuracy by redesigning the mask branch structure.The use of front-and-back layer feature fusion on the mask branch can better preserve the edge information.By increasing the resolution of the RoIAlign layer,more accurate boundary information is obtained.The use of depthwise separable convolution reduces the training parameters without affecting the accuracy of the algorithm,and improves the efficiency of the segmentation algorithm.The experiment demonstrates the accuracy and efficiency of the algorithm by comparing the results of mAP(mean average precision)with the time required to detect quantified pictures and the memory consumed.
引文
[1]Hariharan B,Arbeláez P,Girshick R,et al.Hypercolumns for object segmentation and fine-grained localization[C]//IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2015:447-456.
[2]Dai J,He K,Sun J.Convolutional feature masking for joint object and stuff segmentation[C]//Computer Vision and Pattern Recognition.IEEE,2015:3992-4000.
[3]Pinheiro P O,Collobert R,Dollár P.Learning to segment object candidates[C]//Advances in Neural Information Processing Systems.2015:1990-1998.
[4]Dai J,He K,Sun J.Instance-Aware Semantic Segmentation via Multi-task Network Cascades[C]//Computer Vision and Pattern Recognition.IEEE,2016:3150-3158.
[5]He K,Gkioxari G,Dollár P,et al.Mask R-CNN[C]//IEEE International Conference on Computer Vision.IEEE,2017:2980-2988.
[6]Ren S,He K,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems.MIT Press,2015:91-99.
[7]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition.2016:770-778.
[8]Lin T Y,Dollar P,Girshick R,et al.Feature Pyramid Networks for Object Detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2017:936-944.
[9]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2015:3431-3440.
[10]Girshick R.Fast R-CNN[C]//IEEE International Conference on Computer Vision.IEEE,2015:1440-1448.
[2]Dai J,He K,Sun J.Convolutional feature masking for joint object and stuff segmentation[C]//Computer Vision and Pattern Recognition.IEEE,2015:3992-4000.
[3]Pinheiro P O,Collobert R,Dollár P.Learning to segment object candidates[C]//Advances in Neural Information Processing Systems.2015:1990-1998.
[4]Dai J,He K,Sun J.Instance-Aware Semantic Segmentation via Multi-task Network Cascades[C]//Computer Vision and Pattern Recognition.IEEE,2016:3150-3158.
[5]He K,Gkioxari G,Dollár P,et al.Mask R-CNN[C]//IEEE International Conference on Computer Vision.IEEE,2017:2980-2988.
[6]Ren S,He K,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems.MIT Press,2015:91-99.
[7]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition.2016:770-778.
[8]Lin T Y,Dollar P,Girshick R,et al.Feature Pyramid Networks for Object Detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2017:936-944.
[9]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2015:3431-3440.
[10]Girshick R.Fast R-CNN[C]//IEEE International Conference on Computer Vision.IEEE,2015:1440-1448.