DevlUp Webinar

Deploying Deep Learning Applications

For any deep learning model in general:

• Train the model
• Save the model architecture and weights(json and h5)
• Load up the model and weights and serve them at the needed endpoint

For the MNIST Handwritten App

• Train the model using the given Colab notebook
• Save the json and h5 files
• Load your model through the load.py
• Preprocess your data to store the base64 data into image and further resize and reshape as per the model requirements.
• Send your image as an input to the model
• Store the prediction and make a POST request to the Web App

Steps for running inference app:

Using virtual environment:

• python3 -m venv venv
• source venv/bin/activate (For Linux)
• venv/scripts/activate (Windows)
• pip3 install -r requirements.txt

For starting the app

• python3 app.py