⚠️ MODE BROUILLON SECRET
Intelligence Artificielle
Smart Classification & Industrial Vision
As part of the industry of the future, this project aimed to develop AI models (ResNet-18) for industrial image classification and anomaly detection.
[ 📸 TO INSERT: Screenshot of the web supervision UI ]
The main challenge was tuning hyperparameters against changing lighting conditions (manual white balance required). Once stabilized, the AI was integrated into a real-time Web UI: this interface returned the video feed from the Cognex camera, allowed acknowledgment of production errors, and sent automatic trigger commands to PLCs controlling actuators (e.g., closing a gripper). If a defect was detected, an image was exported via FTP for the quality control department.
- ✓ Tuning neural networks (ResNet-18) to counter false positives and lighting variations.
- ✓ Development of a real-time Web UI interfacing AI and industrial PLCs.
- ✓ Complete Industry 4.0 scenario: Cognex detection, PLC control, and FTP export.
- ✓ [ 📸 TO INSERT: Diagram of the software/hardware architecture ]
PythonIAVision IndustrielleIHM Web