search

cardwing / Codes-for-Lane-Detection

Learning Lightweight Lane Detection CNNs by Self Attention Distillation (ICCV 2019)
MIT License
1.04k stars 333 forks source link

the script which draw the dataset LLAMAS json label to image for 5 classification #318

Closed noobliang closed 3 years ago

noobliang commented 3 years ago

Hi, author, I encountered a problem. How do I get the llamas image label for 5 classification. There hava a script seem to obtain binary segmentation image or multi_class_segmentation_label (l1, l0, r0, r1), whirch are 3-channel image. How to get 1-channel label like the CULane label. Thanks very much.

cardwing commented 3 years ago

You can just use the argmax function to get the index of the maximum activation in each pixel of the multi-class label.

noobliang commented 3 years ago

You can just use the argmax function to get the index of the maximum activation in each pixel of the multi-class label.

thank for your reply

noobliang commented 3 years ago

You can just use the argmax function to get the index of the maximum activation in each pixel of the multi-class label.

作者你好，我还碰到了ERFnet+sad在TuSimple 数据集上精度提不上，一直在95.3% 左右和ENet+sad的96.64%差距有点大，试了deep supervision也没用，而且我训练的是6条车道线。是不是线条的数量也有影响。作者能给什么建议吗。