Food-Detector-Meraki-Snapshot-API-MQTT

A completely customizable Object Detection tool for Meraki camera using Snapshot API with MQTT capabilities.

1. Install pipenv

pip install pipenv

2. In the root directory,

pipenv install

to install the packages and then to use an environment(shell)

pipenv shell

3. Create a classes.json file inside model_dir folder with it's key as your model's prediction id and value as class name.

{"1": "egg", "2": "roti", "3": "rice", "4": "side_dish"}

Here, if the model predicts 1, then egg is the class.

4. Place your very own custom Tensorflow SavedModel folder in model_dir folder. For example, after placing classes.json

model_dir
└── classes.json
└── saved_model
    ├── assets
    ├── saved_model.pb
    └── variables
        ├── variables.data-....
        ├── variables.data-....
        └── variables.index

5. Edit configurations in config.ini file.

• In credential section, enter meraki camera credentials
• In model section, enter saved model folder name. In the above example, model_dir = model_dir/saved_model
• In mqtt section, enter broker and topic details

6. Run

food_detector.py

NOTE: To view what food_detector.py published, run subsrciber.py