The raw data will be processed by generator shell scripts. There will be two subdirs('train' & 'val')
train or val dir {
imageA: contains the A images for train or val.
imageB: contains the B images for train or val.
labelA.json: contains the A json files for train or val.(optional)
labelB.json: contains the B json files for train or val.(optional)
}
The json format for GAN below.
The label json format for Image Classification below.
[
{
"image_path": "imageA/image_name",
"label": class_num,
"bbox": [x_left_up, y_left_up, x_right_bottom, y_right_bottom],
"kpts": [
[x, y, visible],
...
]
},
{
...
}
]