The refined label system is easy to config based on different level content moderation needs for your scenarios
Trained by Massive Data Samples. Continues iteration with hundreds of millions of data, real time content, and suppotrs a variety of content management
Deep integration of Inception, ResNet, MTCNN, InsightFace, EAST, CRNN and other models, multi-model combination training, accurate identification of various illegal pictures, the accuracy rate is as high as 99%
Accurately identify pornographic images such as obscenity, pornographic comics, and child nudity , support pornographic classification, and quickly identify severe/mild pornographic and sexy images
Accurately identify terrorist organizations, cult organizations, guns and knives, bloody riots , child violence and other violent and terrorist pictures in various scenes
Accurately identify the image content of advertisements published using watermark advertisements, personal data screenshots , QR codes , logos , etc.
Accurately identify low-value image, such as low-quality images and negative values
It can identify similar content based on image samples