LadiesMan924
LadiesMan924
> label就是图片的名称啊，你看看自定义dataloader那部分就明白了
I think your learning rate is too big, set it to 0.001 when start training.
> I've tried many times even if the learning rate is set very small, the best accuracy would not be higher than 0.6. Maybe we should use more images for...
I also encountered the same problem. Who knows what the problem is?
I also want to know how to add other features. Are there any results now?
> data_loader中的 num_train = int(len(df_raw)*0.7) num_test = int(len(df_raw)*0.2) num_vali = len(df_raw) - num_train - num_test 三种数据，和seq_len，pred_len有什么关联。 这之间没有什么关联。num_train，num_test, num_vali是对你的数据集进行划分，划分成训练集、测试集和验证集，也就是对原始数据的train_test_split。而seq_len是你在训练和预测时输入的序列长度，pred_len是你希望得到的预测结果的序列长度。
我发现对于拍摄角度比较正的照片经过矫正以后牌照反而变倾斜了，并且对于有的图片倾斜矫正并没有起作用。