Drakflow project is one of the fantastic project, but going in depth of the code I was facing an issue, as darkflow does data augmentation (horizontal flipping) by default, but now I want to add custom shearing augmentation in it.
I have gone though darkflow.net.yolo.predict.preprocess , where I got im as well as allobj variable which has image and annotation information, but over there after doing all operation at the very last step it does im = self.resize_input(im) , which will resize to all images to 413*413 as mentioned in cfg file, but after that only im is return, if I have made modification in allobj variable with new annotation, I am not having idea at the time of training from where does it take annotations for particular image, so can you please tell at the time of training from which file and variable it takes annotation.
Drakflow project is one of the fantastic project, but going in depth of the code I was facing an issue, as darkflow does data augmentation (horizontal flipping) by default, but now I want to add custom shearing augmentation in it. I have gone though darkflow.net.yolo.predict.preprocess , where I got im as well as allobj variable which has image and annotation information, but over there after doing all operation at the very last step it does im = self.resize_input(im) , which will resize to all images to 413*413 as mentioned in cfg file, but after that only im is return, if I have made modification in allobj variable with new annotation, I am not having idea at the time of training from where does it take annotations for particular image, so can you please tell at the time of training from which file and variable it takes annotation.
Thanks & Regards!