search

594422814 / UDT_pytorch

MIT License
192 stars 23 forks source link

About dataset. #7

Closed ReedZyd closed 4 years ago

ReedZyd commented 5 years ago

Is the dataset you used unprocessed after downloading from the official website? The dataset I downloaded from the official website seems to be unable to use because of the label name and the image size. Can you share a complete dataset file? Thank you very much.

594422814 commented 5 years ago

We use the standard VID dataset (video object detection dataset instead of the image detection dataset) without further processing. The only difference is we rename the "train_0000""train_0001"... folders as "a""b""c"... Please check train/dataset/parse.py for more details. You can also refer to DCFNet for the detailed training dataset information.