cfzd
cfzd
@joker-wutong you can refer to `speed_real.py`. https://github.com/cfzd/Ultra-Fast-Lane-Detection/blob/f58fcd5f58511159ebfd06e60c7e221558075703/speed_real.py#L139
@shuizaola 分割网络的作用主要是通过多任务学习来提升性能。差别不大也是有可能的，因为它的作用主要是辅助的。
@98Ray 应该是要把train和test都解压到一个文件夹里面，两个的clips会自动合并。那个clips既包含train的也包含test的
@cxhmikegithub 如果你指的是CULane数据集评估时的IoU的话，这部分的计算是在CULane的评估代码中实现的。具体做法很简单，就是给定GT的车道线点集以及预测的车道线点集，以一个固定的宽度（30个像素）把他们在一个空白画布上画出来，然后计算他们的IoU，具体代码可以在这个地方看到： https://github.com/cfzd/Ultra-Fast-Lane-Detection/blob/f58fcd5f58511159ebfd06e60c7e221558075703/evaluation/culane/src/lane_compare.cpp#L10-L56
@cxhmikegithub 是的，这个是整个数据集的统计量
@ChicagoPark Sorry, we can't provide the res34 model for now. But we will soon release the UFLDv2 work, in which the res34 v2 model could achieve 76.0 F1 on CULane.
@kadirbeytorun It might be a separate repository. And we have already got the review, so you could see it very soon. All models' weights would be shared.
@wuh15 @zailushang-aoh 可以参照这个issue: #8
@bz815 I'm wondering what metrics do you use to evaluate? What are the values of the top1, top2, top3 accuracies during training?
@bz815 It seems the results are pretty good since the top3 accuracy could achieve ~98%. "what metrics" means what kind of source that you get the value of 0.0003? It...