Structure of Datasets

[NhanDoremon]

Dear author, Could I ask you about the structure of your datasets? . I am struggling with training, and also I cannot download the datasets from your link because BAIDU download really slowly with the free version.

[NhanDoremon]

I cannot train the model because of the error when I create label_path from images_path. So that is why I revised the original code from the line 391 in "datasets.py" into the below code, and fortunately It works well:

def img2label_paths(img_paths):
# Define label paths as a function of image paths
sb, t = 'labels', [] # Replace visible or infrared with labels
for x in img_paths:
if 'visible' in x.split(os.sep):
sa = 'visible'
elif 'infrared' in x.split(os.sep):
sa = 'infrared'
else:
print(f"Skipping invalid path: {x}")
continue
# Replace and ensure paths are joined correctly
label_path = x.replace(sa, sb, 1).rsplit('.', 1)[0] + '.txt'
label_path = os.path.normpath(label_path) # Normalize the path
t.append(label_path)
# print("Generated label paths:", t) # Debugging
return t

[jacksonwu09]

你要根据你数据集的结构修改啊