🛡️ SCORE - Sensitive Content Online Reviewer
SCORE is a multi-label sensitive content detection API designed to identify a wide range of explicit, harmful, and policy-violating content in images. Unlike traditional NSFW detectors that only flag nudity, SCORE provides advanced multi-class classification tailored for modern content moderation needs.
🔍 Key Features
- Multi-class classification — detect various categories beyond nudity
- Lightweight & Fast — suitable for integration into websites, apps, or moderation pipelines
- Easy-to-parse JSON output — returns class labels with confidence scores
🧠 Supported Classes
Class Name | Description |
---|---|
negative | Generally poor-quality or corrupted images |
nudity | Explicit male/female nudity |
anime_porn | Hentai or anime-style pornography |
gore | Violence, wounds, blood, or extreme injury |
firearms | Guns, rifles, or similar weapons |
bladed_weapon | Knives, swords, or other bladed weapons |
low_alcoholic_beverage | Alcoholic drinks under 5% (e.g., beer, IPA) |
high_alcoholic_beverage | Alcohol over 5% (e.g., wine, vodka, liquor) |
drugs | Cannabis, marijuana, ecstasy, pills, etc. |
🆚 Compared to Traditional NSFW Detectors
Feature | SCORE | NSFW Detector |
---|---|---|
Accuracy (Nudity) | ~88 - 90% | ~95 - 97% |
Multi-class Detection | ✅ Yes (10+ sensitive categories) | ❌ No |
Firearms / Drugs | ✅ Yes | ❌ No |
Output Format | ✅ JSON with confidence scores | ✅ Simple binary/score |
Use Case Scope | Wide (e.g., weapons, alcohol) | Limited (NSFW only) |
While NSFW detectors offer slightly higher accuracy for nudity, SCORE delivers a much broader range of sensitive content detection, making it more versatile for real-world applications.
💡 Use Cases
- 📱 Content moderation on social media platforms
- 🛒 Product image filtering on marketplaces
- 🎮 Safe community management for games and avatars
- 🧠 Pre-filtering visual content for LLMs or AI apps
- 🧑💼 Visual content compliance monitoring in workplaces or education