Image classification using Python/TensorFlow -- 2

This project contains two modules, one is training set of images and second one is predict 1. Training: program should take set of images to train with one label, and if i give more images in future, i should able to retrain with new set of images on top of old model of same label. Ex: i trained 10 cat images first time, in future, i should able to train 5 more cat images on top of Cat model, (Note: I tried this using TensorFlow, but when i retrained with new set of categories, old categories are getting removed, we should keep old categories as it is, and retrain new categories on top of it) 2. Predict: I will send image to program, it has to predict which category it belonging too. Example: step 1: I have 2 categories in dataset folder (cat, dog) thos categories folders having images step2: i trained that dataset step 3: tensorflow will generate two files ( [login to view URL], [login to view URL]) step 4: we can predict with those files step 5: now i removed (cat, dog) folders from dataset folder, because i already trained, so no longer needed these images step 6: i will add two more categories in dataset folder (tiger, snake) User Avatar step7: i need to retrain the model, because i added two more categories step 8: when again new cat images came, i should able to retain again with these new dataset on top of existed model< Everyday 50 plus categories(dataset/categories) will be added, so every time i can not retain whole dataset, i should able to train categories and remove those images as when it comes, so it takes less time to compile new categories, and model should able to predict/classify on old & new Useful links: [login to view URL]