PMLDL MLOps assignment

A project to showcase model deployment using Flask and Streamlit

Problem

I chose an image classification problem for this task. Specifically, I used CIFAR-100 dataset to train Effnet-b2 model. Due to computational limitations, I only trained for 10 epochs. However, I managed to achieve accuracy of 0.88.

Dataset used

CIFAR-100 (course classes)

Model used

Effnet-b2

Metrics

Accuracy: 0.88

Demo Streamlit app

Flask backend routes

How to start

Everything runs via docker compose

docker compose up 

Streamlit

Got localhost:3000 for Streamlit

Flask

Currenly, only one route is avaliable

POST /classify

Here's an example curl command

curl --location 'localhost:8000/classify' \
--form 'image=@"/path/to/img.png"'

Any PIL compatable image works

Output in the class of the image