Scanner for illegal content on vichan based imageboards
The concept here is to create an AI bot which can read images and posts on imageboards, then send out notificaitons and perform actions if content is illegal. This is targetting "for proffit" spammers who use imageboards to post CSAM content and intend for people to pay for it.
This is different from CSAM spam which intends to shut down websites through complaints to registrars or providers. It is a means to target people spamming for money.
It's javascript because I'm too lazy to setup a ts and webpack build. Requires an outside server using OCR and a disposable webserver hosting it(will be experimenting with shared hosting on A2Hosting( https://www.a2hosting.com/web-hosting/)
since they seem to offer a good and cheap NodeJS solution for some lightweight webscraping. Is designed for vichan imageboards but with little effort can retarget other platforms.
Though this is no substitute for good moderation practices and 'preventative maintenance' it can help deal with deadspots in moderation and make it so that spammers must exhaust even more energy to achieve the goal of for proffit spamming.
Uses vichan but others possible
Designed to run on shared hosting
Will add details of how kissu approaches OCR for reaiding watermarked images. Some current ideas are https://github.com/PaddlePaddle/PaddleOCR and https://github.com/zhang0jhon/AttentionOCR which are high quality scene text recognition solutions.