Automate extraction of study notes from WhatsApp Images

We all end up with hell lot of images to be deleted at the end of each semester. I've trained a CNN model to predict such images and extract them out of WhatsApp Images directory

Images dataset can be found on Google Drive

like this:

Requirements:

• Numpy
• Keras
• TendorFlow

Instructions:

• Download and extract the project folder.Open the terminal in the extracted folder and Install dependencies using pip install -r requirements.txt.
• Connect your Smartphone to your system(Linux/Ubuntu OS), mount Internal Storage and copy the absolute path to the WhatsApp folder, to know the absolute path open a terminal in WhatsApp folder and run pwd command(in Linux OS) or copy the location by choosing properties from right-click menu (in Windows OS) .
• Run the extract.py script by python extract.py and paste the copied path when asked to. The script will create a new folder named notes in your WhatsApp Image folder and move the study notes images to it.

We've trained the model on about 1000 images and using Keras' data augmentation pipeline. Currently the model is 92% accurate on my dataset. Feel free to add your own data and train the model on it to make the model more accurate. To add your own data, create a data folder in behind_the_scenes folder, create two subfolders 1 and 0 inside data, in 1 put study notes and put all other important images in 0. See behind_the_scenes folder for more info.