Generate COCO style dataset

After downloading and extracting the data, we need to convert it to a format that the network expects. We’ll do that now.

Remember where you extracted your archive. For this tutorial, we’ll consider ~/Desktop/TrashBot as the extracted location.

Now, inside the root directory of the project, make the build_dataset bash file an executable. On Linux, you can do it like this:

chmod u+x build_dataset.sh

Now, go ahead and run the build_dataset.sh script as follows:

./build_dataset.sh --data_path ~/Desktop/TrashBot --output ~/Desktop/coco

where –output represents where I want the COCO dataset to be stored on disk and –data_path represents my Supervise.ly extracted archive

That’s all! You’re all set. Now, go ahead and train a model.