Q/Anons:
MilVet/EE here.
Some hopefully helpful tips for image recognition defeat.
Attached image is expressly designed to overload any image recognition program by essentially creating too many data points and randomized noise to be analyzed.
Reference article here:
https://m.slashdot.org/story/335787
Use a couple instances of this image in your memes in random locations.
Handwriting is really hard for computers to read. Think "Captchas", especially with overlapping lines and slanted writing.
Use software to skew/bubble text and images to create depth. Basically, make things as non rectangular as possible.
Q, what about setting up a counter-image processing "stegosploit"? Very complex to code, beyond my skills, but could be very effective. Implemented as a user friendly app.
https://github.com/csh/stegosploit
Also worth researching: Generative Adversarial Networks. Good for making deepfakes, why not memes?
Patriots fight!
Against all enemies, foreign and domestic.
"As the President may direct".
Here am I, send me.