chengKun
  • Joined on Dec 01, 2022
Loading Heatmap…

chengKun closed issue chengKun/1-18-2#3

1-18-2U-net车道检测

2 weeks ago

chengKun commented on issue chengKun/1-18-2#3

1-18-2U-net车道检测

车道线检测方法分为两种:一种是传统算法,主要基于边缘特征或者是图像分割,通过图像预处理,特征提取等方式与卡尔曼滤波器等算法结合,在识别出车道线后采用后处理的方式形成最终的车道。然而这种传统方法易受到光照变化,行驶车辆,道路破损等干扰,导致效果欠佳。近年来,深度学习方法利用网络模型自动学习目标特征,具有较强的泛化能力,可以有效提高目标检测的准确率。得益于卷积神经网络的强大特征提取能力,性能也在不断提升。

2 weeks ago

chengKun commented on issue chengKun/1-18-2#3

1-18-2U-net车道检测

H-Net比用固定的转换矩阵要好,用3阶多项式拟合要比2阶好,用了透视变换比不用好。 车道线识别深度学习面临的挑战: (1)车道线这种细长的形态结构,需要更加强大的高低层次特征融合,来同时获取全局的空间结构关系,和细节处的定位精度。 (2)车道线的形态有很多不确定性,比如被遮挡,磨损,以及道路变化时本身的不连续性。需要网络针对这些情况有较强的推测能力。 (3)车辆的偏离或换道过程会产生自车所在车道的切换,车道线也会发生左/右线的切换。一些提前给车道线赋值固定序号的方法,在换道过程中会产生歧义的情况。

2 weeks ago

chengKun opened issue chengKun/1-18-2#3

1-18-2U-net车道检测

2 weeks ago

chengKun created new model 1-18-2_model_9tvx

2 weeks ago

chengKun created CPU/GPU training task cheng202301182046384

2 weeks ago

chengKun upload dataset label.png

2 weeks ago

chengKun upload dataset label_viz.png

2 weeks ago

chengKun upload dataset label_names.txt

2 weeks ago

chengKun upload dataset img.png

2 weeks ago

chengKun upload dataset label_names.txt

2 weeks ago

chengKun upload dataset label_viz.png

2 weeks ago

chengKun upload dataset label.png

2 weeks ago

chengKun upload dataset img.png

2 weeks ago

chengKun merged pull request chengKun/1-18-2#2

上传文件至 'tra'

2 weeks ago

chengKun pushed to master at chengKun/1-18-2

  • 3443ef23da Merge pull request '上传文件至 'tra'' (#2) from devtrain into master Reviewed-on: https://openi.pcl.ac.cn/chengKun/1-18-2/pulls/2
  • 7e5201944c 上传文件至 'tra'
  • Compare 2 commits »

2 weeks ago

chengKun created pull request chengKun/1-18-2#2

上传文件至 'tra'

2 weeks ago

chengKun pushed to devtrain at chengKun/1-18-2

2 weeks ago

chengKun merged pull request chengKun/1-18-2#1

上传文件至 ''

2 weeks ago

chengKun pushed to master at chengKun/1-18-2

2 weeks ago