Apppublic

swastikdl/siemens-energy-welding-inspection

sourceHugging Facemitupdated 7mo agoView on Hugging Face

0likes

App README

Siemens Energy: Welding Inspection Portal

AI-powered welding defect detection using YOLOv8. Upload welding images to automatically detect and classify 6 types of defects:

—Lump defect — Excess material buildup
—Spatter defect — Metal droplet splashes
—Pin hole defect — Small holes in the weld
—Chips & Burr — Rough edges and fragments
—Undercut defect — Groove along the weld toe
—Welding protrusion — Excess weld metal protruding

Features

—Batch image upload and live inference
—Adjustable confidence threshold
—Analytics dashboard with defect distribution charts
—Inspection history with image viewer
—CSV report export

Local Setup

bash

pip install -r requirements.txt
streamlit run app.py

Model

YOLOv8 Medium trained on welding defect dataset with frozen backbone (layers 0-7), early stopping, and controlled augmentation. See model_training/ for training scripts.