AI-Powered Inmate Emotion & Mental Health Analyzer
Work Summary
An inmate mental-health assessment system that combines facial emotion recognition, counsellor surveys and document understanding, then uses an LLM to classify mental-health levels.
The Challenge
Assessing the mental health of inmates is slow and depends on limited counsellor time. The goal was to detect emotional states automatically and combine them with existing records to support counsellors.
- YOLOv11n-based facial emotion recognition trained on FER2013 to detect emotional states automatically.
- Counsellor and in-charge survey inputs combined with RAG document understanding (chunking, embeddings, vector DB).
- LLM-based classification of mental-health levels.
- Personalized activity recommendations and an automated analysis report.
The Solution
The analyzer gives counsellors an automated first assessment and a clear report for each inmate, so they can focus their time where it is needed most.
Technologies
PythonYOLOv11nOpenCVDeep LearningLLMsRAGVector DBs
