Video Input
Raw video ingestion.
AI SystemEngineering Work
DCRE MVP
Multi-model AI pipeline analyzing video frames across 22 safety categories with explainability.
~0.5–1s per frame processing with multi-user support.
AIComputer VisionExplainability
System Architecture
Diagram focused on system boundaries, data flow, and integration responsibilities.
DCRE AI System Pipeline Diagram
Frame Extraction
OpenCV
Frame sampling & preprocessing.
Multi-Model Inference
PyTorch, Transformers
Parallel model execution.
Classification Layer
22 safety categories classification.
Aggregation Engine
Combines multi-model outputs into structured result.
Results Storage
MongoDB
Stores results + metadata.
Visualization Dashboard
Streamlit
Visualization + inspection.
~0.5–1.0s per frame~0.3–1 FPS processingBatch + semi-real-time capable