The chatGPT-era
Firstly, I am in love with AI now, it can do so much 🤩. In just a matter of a few weeks/months, I learned EfficientNets very sincerely & fell in love with them LITERALLY.
As part of deepfake detection techniques, I used EfficientNet-B4 & B7. Trained both of them with the latest deepfake datasets & measured their detection performance. It was phenomenal to see the results getting generated & seeing all the 0s & 1s. Once I had the results of the research questions, I started writing chapter 4 of the dissertation.
MOST EXCITING PART OF IT ALL.
In the first phase of the research, we had untrained EfficientNet-B4 & B7 as if they are in a real-time setting, where machine learning models get exposed to deepfake images or videos & how well can models detect deepfakes!?
…….. Without numbers, no theory makes sense. (typical #STEM brain🤓)
Even without training, EfficientNet-B4 & B7 models were getting the hang of deepfake detection. Results of the first phase were promising. After getting the results, I used those numbers in McNemar’s test.
In literature-y lingo, it means our p-value surpassed the assumed threshold (⍺=0.05), the study cannot reject the null hypothesis and assumes no significant difference in the effectiveness of EfficientNet-B4 & B7.
Phew… Finally, some data to measure the effectiveness of machine learning models. Next is having trained EfficientNet-B4 & B7 to compete against each other. 🤼♀️ Stay tuned.
some brain dump to get you up to speed!! In a sapiosexual way & the journey itself ofc! by Jasmin Bharadiya
