search

microsoft / VoTT

Visual Object Tagging Tool: An electron app for building end to end Object Detection Models from Images and Videos.
MIT License
4.27k stars 833 forks source link

Add data augmentation feature #973

Closed wwdok closed 4 years ago

wwdok commented 4 years ago

Hello, VOTT's developers ! I hope that VoTT could have a feature of image augmentation when exporting tfrecord, because as a CV amateur, i can only collect limited images, but i also want sufficient data to train the model well, so i expect there will be a feature of image augmentation, that is to say, in the Export Settings UI, after users choose tensorflow records Provider, then users can choose any type of image augmentation they like (e.g. flip, saturation, brightness) and adjust their parameters, finally save export settings. By the way, can we set the numbers of exported tfrecords, because if one image one tfrecord, it look like a little verbose and sometimes unconvinient.

wwdok commented 4 years ago

well, today i find there has already been data augmention configuratioon at config file, sorry for my ignorance