AI & CV Powered Verification Engine

can a camera check whether a college has the infrastructure it claims?

We explored whether computer vision could assist parts of AICTE infrastructure verification, using plans, photographs, documents and inspection video.

blueprints + photos + documents + video → assessment signals

several pipelines. several open questions.

Floor-plan segmentation and OCR estimate rooms and dimensions. YOLOv8 with SAHI sliced inference explores equipment detection and workstation counting.

Document checks use OCR, deskewing and blur detection. FFT and moiré analysis explore suspicious screen re-recordings in video. These are experimental assessment signals, not verified compliance decisions.

python · FastAPI · OpenCV · numpy · pytorch · YOLOv8 · SAHI · EasyOCR · FFT