I have a image processing deep learning api built on Python 3.7, Keras, Keras-Retinanet and Flask which basically detect object and then predict it's brand. Before training the model currently we manually prepare the dataset. The dataset contains many brands folders having their one product images. Currently I have 30 brands and this will grow with times. So manually preparing dataset is not a scalable solution. Therefore I want to create a Python script with flask/django UI which can automate this manual process. Current steps followed in project which need to be automated are-
1. Prepare dataset:
a. Inside 'dataset' folder create three folders- 'train', 'test', 'val'
b. Inside 'train', 'test', 'val' folders create a new folder with brand name
c. In newly created folder put cropped images of that brand in the ratio of 80:10:10
2. Start training of the model
I want to automate the first step. Means a user will upload the actual image -> desired object will be detected -> object will be cropped from that image -> detect the brand names from cropped images -> create a new folder with brand name if not exist in 'train','test','val' folders -> put cropped images to 'train','test','val' folders as per their brand name -> start the training script.