Detection Model Training Task using VOC

[zhangyimeng960916]

When using the VOC dataset (such as VOC2012) for the target detection model training task, do I need to convert the VOC dataset into a specific format when parsing the VOC dataset in FATE?

As I know, directory of VOC12 dataset :

• Annotations
• ImageSets
• JPEGImages
• SegmentationClass
• SegmentationObject

is it needed to be converted to the following format (eg. MNIST):

• images
• config.yaml
• filenames
• targets

how to create a target detection training task（pytorch backend）?