RESEARCH INTERNSHIP / INDUSTRIAL COMPUTER VISION
Problem
Detect surface defects in industrial imagery and move predictions through a usable inspection pipeline.
Approach
An object-detection pipeline connected through an EC2 endpoint, FastAPI service, Raspberry Pi, and persistent storage.
Results
Reported mean average precision improved from 56% to 78% in the project evaluation.
My contribution
ML research internship developing surface-defect detection and optimising the detection pipeline.
01Industrial imagery
02Defect detection
03Inspection output
Project account: Build, Learn, Repeat. Metrics describe the internship evaluation, not a general benchmark.
All research