Ram Maheshwari Logo Image
Mohammad Ausaf

औषधHub

This End-to-End Machine Learning Project is designed to tackle the challenge of identifying medicinal herbs. This application employs a fine-tuned ResNet model using the Mendeley Indian Medicinal dataset, achieving an impressive testing accuracy of 98%.
The training and testing dataset contains over 1500 images across 30 medicinal herb species

Project Image

Project Overview

Medicinal plants are the source of raw herbal medicines. They are less costly and appear to produce less undesirable side effects than modern medicines. There are approximately 7,200 medicinal plants known in India, of which hundreds of plants have medicinal properties in their leaves. Identifying such valuable plants often requires an expert or a manual. Therefore, there is a need to automate processes to retrieve information more quickly.

Following is a loose architecture of the project :

Project Image

The Project is divided into two major components:

1. A client side Front-End, The user uploads the image on front-end, which is sent to the Back-End via a POST request over API.

2. A Model hosted on Google Cloud in a Flask-App which performs Pre-processing on user Input before Inferencing and returning the result.

I invite you to take a look at the GitHub repo linked below for more details, hope you will like it!
Oh! i almost forgot, If you are interesetd in developing this project further, or you have any input for me, maybe an Idea, or even better, a critique, go ahead shoot me a DM or mail, or both (if you are restless like me!).
Untill we meet again, on next project page ofcourse! 🫡

Tools Used

Python
NumPy
Pandas
PyTorch
OpenCV
Flask(WebApp)
HTML
CSS
Google Cloud