TCM Hob Inspector

General information

Multifunctional system for the automatic inspection of gear hobbing tools, featuring:

• AI-based image processing system
• Dash based interactive application
• SQL database for storing the results of the performed scans

Machine learning solutions

• Two stage inference with YOLOv8 object detection model and Segment Anything Model (SAM)
• Supervised learning using custom datasets
• Annotations with artifacts saved in .json format, compatible with LabelMe software
• System for semi-automatic datasets creation
• Custom python scripts for training and evaluation of AI models
• Model comparison and performance tracking using Neptune.AI

Technologies used

• Python 3.11
• YOLOv8
• PyTorch 2.0
• NVIDIA CUDA 11.8

Features

Hob wear map.

Automatic report generation.

Organisations involved in the project development

Wrocław University of Science and Technology	Machine Vision Laboratory	TCM International