Submission for Precog Recruitment Task 2: Analyzing Hateful Memes
This codebase uses various computer vision techniques to qualitatively and quantitatively analyze the different aspects that include and revolve around hateful memes. The analysis gives a very interesting insight into patterns, context and visual cues, all of which affect the way we pervieve memes. The project also highlights the difficulties faced in idenitfying whether a meme can be classified as hateful or not.
The codebase has been divided into folders corresponding to the subtasks of the main task.
IMPORTANT Before you begin, you should download the Dataset from this (Google Drive Link) and extract it into the main folder. If you wish to use your own dataset, feel free to do so, but make sure to follow the same directory setup as the one in the link.
All the codes and outputs related to this task can be found in the object Detection folder.
All the codes and outputs related to this task can be found in the Caption Impact Assessment folder.
NOTE The notebook when compiled and run, outputs the meme images with their text/caption removed. In my submission, I have demonstrated the model's results by running it on this dataset. The output for the same can be viewed in this Google Drive Link
All the codes and outputs related to this task can be found in the Classification System Development folder.
All the codes and outputs related to this task can be found in the Bonus Task folder.
The PDF Report for the paper reading task can be found here.
The following libraries need to be installed in order to run this submission
Python 3.10 or above
cv2
pytesseract
torch
os
string
numpy
keras_ocr
tensorflow
matplotlib
Run the following command in your terminal to install the required dependencies:
pip install - matplotlib os torch cv2 numpy keras_ocr pyterrseract tensorflow string