Weld Defect Classification System

Chirag Patankar

Completed work

AI Engineer

PyTorch

Streamlit

Manufacturing

Weld Defect Classification

Overview

This project is a deep learning-based system for classifying weld defects using convolutional neural networks (CNNs). The system processes weld images and predicts whether a weld is good or defective, categorizing defects into six different classes.

Features

Trainable Model: Uses PyTorch to train a CNN on labeled weld images.

Testing Module: Evaluates model accuracy using test images.

Web Interface: A Streamlit-powered UI for real-time image classification.

Binary and Multi-Class Classification: Can classify welds as Good or Defective, and further categorize defects into specific types.

Installation

Prerequisites

Ensure you have Python 3.8+ installed. You also need the following dependencies:

pip install torch torchvision streamlit numpy pandas matplotlib pillow

Dataset

The dataset used for training and testing is available on Google Drive. You can download it using the link below:

Download Dataset

Usage

Training the Model

To train the model, run:

python train.py

This will load training images, process them, and train a CNN model. The trained model is saved as modelv5.pt.

Testing the Model

After training, evaluate the model using:

python test.py

This script loads test images and prints evaluation metrics such as accuracy.

Running the Web App

To launch the Streamlit UI for classifying weld images, run:

streamlit run app.py

This will start a web-based interface where users can upload images and receive predictions.

Project Structure

├── data/                  # Directory for training and test datasets
├── train.py               # Training script
├── test.py                # Testing script
├── app.py                 # Streamlit UI for classification
├── modelv5.pt             # Trained model file (generated after training)
└── README.md              # Project documentation

Model Details

CNN architecture with multiple convolutional, batch normalization, dropout, and ReLU layers.

Uses cross-entropy loss and Adam optimizer for training.

Outputs a classification prediction with probability scores.

Convolutional Neural Networks (CNNs)

Convolutional Neural Networks (CNNs) are a class of deep learning models specifically designed for image processing tasks. They consist of multiple layers, including:

Convolutional Layers: Extract spatial features from images by applying filters.

Pooling Layers: Reduce dimensionality and retain essential information.

Fully Connected Layers: Combine extracted features to make classification decisions.

Activation Functions (e.g., ReLU): Introduce non-linearity to improve learning capability.

CNNs are effective for weld defect classification as they can learn to recognize intricate patterns in weld images, distinguishing between different defect types accurately.