oyoanan

images

6

您好，images下载路径无法打开。请问是BDD数据集中的\BDD100K\bdd100k_images\bdd100k\images\100k下的train和val吗？

test_label_classes.txt

6

Hello, is there an update to test_label_classes.txt? The error occurred due to incomplete correspondence between the test-label_classes.txt downloaded from the link you provided and test-label.json. For example, 5 lane lines...

dataset

1

Hello, the dataset folder structure downloaded from the Tusimple official website includes train_set, test_set, test_label.json, and test_label.new.json. Is it necessary to merge train_set and test_set for dataset creation in your...

Thank you for your work. I only saw the category labels of the training set. Do you have labels for the testing set?

你好，论文中显示TUSIMPLE数据集的Validation358张，但是代码运行显示每次训练时的验证似乎是 2782张（Test）。这是怎么回事呢？

作者您好，我在其他项目中加入您的车道线训练标签过程中，似乎因为DALI不支持多进程导致数据增强失败。所以，想问下您的车道线标签数据，是只能使用DALI增强，还是可以使用其他类似的库吗？最重要的是这里的labels要如何跟随图像的变换而变换呢？希望可以得到您的回复，非常感谢！ ` def ExternalSourceTrainPipeline(batch_size, num_threads, device_id, external_data, train_width, train_height, top_crop, normalize_image_scale = False, nscale_w = None, nscale_h = None): pipe = Pipeline(batch_size, num_threads, device_id) with pipe: jpegs, seg_images, labels =...

Hello, is the category file for the training set not provided?Looking forward to your reply,Thanks!!

您好，为什么这里4条车道线都在行列锚上进行了预测，而不是像论文中2条行锚、2条侧锚呢？希望得到您的回复感谢~ pred_dict = {'loc_row': out[:,:self.dim1].view(-1,self.num_grid_row, self.num_cls_row, self.num_lane_on_row), 'loc_col': out[:,self.dim1:self.dim1+self.dim2].view(-1, self.num_grid_col, self.num_cls_col, self.num_lane_on_col), 'exist_row': out[:,self.dim1+self.dim2:self.dim1+self.dim2+self.dim3].view(-1, 2, self.num_cls_row, self.num_lane_on_row), 'exist_col': out[:,-self.dim4:].view(-1, 2, self.num_cls_col, self.num_lane_on_col), }