Implementation of two simple keras models for automatic image quality assessment and product recognition in images (here, watch/clock recognition). The models are further exposed in an API (Flask).
Install all requirements. (N.B.: python 3.6.3)
pip install -r requirements.txt
You may follow all development steps and questions in the dev_model.ipynb notebook. (also in .html). However, as the file is quiet big, you would need to clone this repo to visualize the notebook.
you may have information on how to run predict.py by following:
python predict.py --help
You just need to store all your images (.jpg, .jpeg, .png) in one unique folder and run predict.py by following the help instructions.
└── olx_test
├── app # data folder contaning both A and B images
├── models # folders where .h5 keras model files are stored
├── app.py # the flask app
├── Procfile # Procfile necessary for deploying on Heroku
├── requirements.txt # all requirements necessary for the flask app
└── runtime.txt # version of python (3.6.3) for deploying on Heroku
├── dev_model.html # The html version of the notebook
├── dev_model.ipynb # The dev notebook
├── pedict.py # .py to deploy on your own images
└── utils.py # utils (functions call in the notebook)