Closed Jail0910 closed 4 years ago
You can either implement a data loader for you dataset, similiar to the TuSimple one provided, or transform your annotations to the TuSimple format, described here.
In the one provided, each JSON here will be loaded. Their paths will be joint with the dataset root path. That is, if you pass the argument root='tusimple'
and the JSON path is train.json
, it will load the file in tusimple/train.json
and read the annotations in the aforementioned TuSimple format.
You can either implement a data loader for you dataset, similiar to the TuSimple one provided, or transform your annotations to the TuSimple format, described here.
In the one provided, each JSON here will be loaded. Their paths will be joint with the dataset root path. That is, if you pass the argument
root='tusimple'
and the JSON path istrain.json
, it will load the file intusimple/train.json
and read the annotations in the aforementioned TuSimple format.
thank u, I have settled it
@Jail0910 how does the transformation to .json
format works? I've read the linked links, but struggle with the implementation idea.
如何准备用于模型训练的标签? 您好,我刚开始学习车道线检测,请问这个怎么解决的?
How to prepare the label for model training?