Skip to content

Latest commit

 

History

History
51 lines (33 loc) · 2.05 KB

2019-05-第二周.md

File metadata and controls

51 lines (33 loc) · 2.05 KB

2019-05-第二周

标签(空格分隔): CASIA


工作总结与安排

上周工作

  • 使用MRCNN模型提取有用信息;

    image_1db2ghujiivcis11j2g1cj1jqv6p.png-134kB

    使用MRCNN模型对图片做预测,下面这个数字是使用一块GTX 1080跑一天处理的图片数(大概一秒1张半的样子);这个方法速度慢,但是基本上每天都可以有结果;

    image_1db2gb0r41a18d51mau13aabcp3e.png-162.2kB

  • 使用SSD进行目标检测;

    image_1db2gj5p5vm8e1i05mbf1ms17m.png-60.1kB

    下面这个结果是116/2565,大概可以检测到的信息比率是4.5%,约20ms处理一张;SSD的处理速度较快,但是有大部分图片会没有信息,有信息的话绝大多数是单目标;

    image_1db2ghc5e15e0i3m13v74ph1j546c.png-101.9kB

下周安排

图像分割 Fully Convolutional Networks for Semantic Segmentation (FCN) Mask R-CNN Fully Convolutional Instance-aware Semantic Segmentation(FCIS) FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation Learning Deconvolution Network for Semantic Segmentation Learning a Discriminative Feature Network for Semantic Segmentation

点云相关 Stereo R-CNN based 3D Object Detection for Autonomous Driving PointRCNN: 3D Object Proposal Generation and Detection from Point Cloud Escape from Cells: Deep Kd-Networks for the Recognition of 3D Point Cloud Models

图卷积 SEMI-SUPERVISED CLASSIFICATION WITH GRAPH CONVOLUTIONAL NETWORKS Learning Convolutional Neural Networks for Graphs