Megvii-BaseDetection/cvpods

About test model in train mode

www516717402 opened this issue · 4 comments

Suggest add a small data for train code since prepare huge dataset will get error somtimes and small dataset for comrehend this detect structure quickly. Like Yolov5 coco 128 images.

If I understand what you mean correctly, we can provide an example using torchvision's CIFAR10 to train a classification model.

Does it satisfy your expectation?

About detection Please offer a small dataset(such as 128 images subset in coco and this dataset put in google drive or baidu disk for download) for run cvpods rapidly.

The above are all my thought. Close this issue.

we think providing a mini-dataset is not necessary. Users should prepare the dataset by themselves, not by us.