Image Classifier and Counter

Stock keeping using machine vision and database program. For referral, see something like this guy were doing: [login to view URL] Use tensorflow or any ML frameworks that can be trained to recognize standard publication books spine (training data for the model will include a set of 100-200 image of unique individual physical books, viewed when it is laid on it's side, and labeled title/or whichever convenient stock keeping unit/barcode etc) within very controlled environment (webcam will take test (data input) picture in very organized way with yellow background). Bounding label box is not needed, realtime detection not needed, any other bells and whistle is not needed. Anything will be done to increase probability of detection and speed. We think it will be used like a cashier checkout barcode scanner at a bookstore. So less worrying about any messy stuff in the input data. But instead scanning books one by one, it snaps the book spine all at once, on a very neat and visually appealing background, like a on a standardized tray. the camera or webcam will be fixed distant from the image capture area. We might buy and use something like this: [login to view URL] We want the program to be able to recognize the still image, count how many instances of the book in the still picture and yield the output (prediction). This output will include [Title] [ISBN/SKU#] [Price] [Qtty] [Total] and finally [Grand Total] of all the book inside the still image. Maybe need [Confident Interval too]. Output is preferred to be in both PDF and also .csv or suitable format that can be further processed in MS Access. Maybe we will output into something like a store receipt. The important thing is; the model must be able to be reset and updated with sets of training data at will on daily basis, (think everyday the bookstore may have new unique book title in it's inventory everyday) it's okay to train it for 8 hours everyday before store opening. Ability to correct prediction output and feedback to the model is a plus. The guy from the youtube; the file is quite messy but we dont mind, heck as long as it works. But if you can put everything neatly that will be great too.