Visual Information Structuring
Pedestrian Feature Extraction
• Support various recognitions including facial feature, gender, age range, dressing color, glasses, backpack, etc.
• Support uniform recognition especially for features such as religious dress(black robe, etc.) and weaponry（rifle , sword, etc.）
Vehicle Feature Extraction
• 200+ Major brands: 95%; 3000+ models: 90%
• Partial Feature: 90%
• Searching by image; similarity of the first 100 images is 85%
• Based on deep learning, industry leading algorithm
Pornographic, politically sensitive characters, terrorists and other illegal image content could be accurately identified through image recognition, face recognition, feature recognition, natural language analysis and other technologies. It could customize character filtration, texts filtration such as comments, and improve efficiency of filtering video content to saves 99% of the labor cost.
By the means of searching picture, audio and video, the specified image/video fragments or the image/video files with the highest similarity could be directly located in the massive multimedia database.