Implemented Machine Learning Algorithms

Deployed a Flutter web application connected to a machine learning API I made with Python, and Docker!

The web application can predict food images based on their URL link, and is deployed to Heroku using Node.Js

You can check the app here: https://food-vision-ruaro.herokuapp.com/

I developed a machine learning model that can classify food images and deployed it as a RESTful API service using FastAPI, Docker and Heroku

Completed my summer research internship on use of CNN-LSTM in buried explosive object detection.

Currently working on my own algorithm to improve model performance.

Worked on selecting and optimizing machine learning algorithms for mar-tech lead prediction. Successfully provided a best classification algorithm with 92% accuracy. The task was to create a POC for creating leads for the sales team of client from 5+ million records.

Text Classification with BERT: A beginner's Guide to fine-tuning BERT
I created a detailed kernel on fine-tuning BERT for text classification. It also acts as a tutorial for experiment tracking and TensorFlow data API. 

https://www.kaggle.com/au1206/fine-tuning-bert-text-classification

A realtime sudoku solver using OpenCV and Keras. It is really a fun project.

The tools I used to build this
1. #OpenCV to capture the image and find the 4 contour of the grid and cells. Also for image processing to prevent overfit and finally to display the result.
2. Subset of the Chars74K dataset to train my model. (I have another project with MNIST dataset.)
3. #Tensorflow and #Keras to train the model to recognize the (hand written) digits.
4. Backtracking algorithm to solve the sudoku grid which I shamelessly copied from the internet.

#OpenCv #computervision #tensorflow #keras #python3 #machinelearning #datascience