Detecting Defects in Steel Sheets with Computer Vision

Go to main | Course Page

Section 1: Introduction

  1. Defect Detection in Manufacturing
  2. Data Overview

Section 2: U-Net Model Architecture

  1. Encoder: Going Down the U-Net
  2. Decoder: Moving Up the U-Net

Section 3: Evaluation Metrics

  1. What is Dice
  2. Cross Entropy vs. Dice
  3. Setup Environment and Import Data

Section 4: Setup, Helper Functions and Explore Data (Code Demo)

  1. Helper Functions for Encoding and Decoding Masks with OpenCV
  2. Helper Function for Dice Loss Calculation
  3. Set Device Configuration
  4. Exploring Data

Section 5: Train Model with BCE Logits Loss (Code Demo)

  1. Define Hyperparameters
  2. Initialize Dataloader
  3. Initialize UNet Model
  4. Define Loss and Optimizer
  5. Train Model
  6. Evaluate Model
  7. Check performance when training with entire data

Section 6: Train Model with BCE Logits Loss and Learning Rate Scheduler (Code Demo)

  1. Train Model with BCE Logits Loss and Learning Rate Scheduler (Code Demo)

Section 7: Train Model with Dice Loss and Learning Rate Scheduler (Code Demo)

  1. Train Model with Dice Loss and Learning Rate Scheduler (Code Demo)

Section 8: Visualize with TensorBoard (Code Demo)

  1. Initial Setup
  2. Initialize Writer
  3. Show Images in TensorBoard
  4. Show Model Architecture in TensorBoard
Report abuse