Work has been done with RedEyesBlueDragon (Suat Tarık Demirel)
For more information about the project, read fuzzer_report.pdf 
To summarize, we have written an ML model to predict a fuzzed file.

fuzzer.cpp:
readFile() function is used to generate fuzzed data from the selected base image (currently, the image "tmp.jpg" is used)
dataset is generated to the folder "dataset".

text.py:
After running the fuzzer.cpp and creating dataset, running text.py script would generate predictions in the "predicts" folder, after training is finished (training uses "testing/ia_100000106.jpg" file as base image).