Computer Vision using Deep Learning - Part 1 & 2
Tue 15 Apr 2025 10:00 AM - Wed 16 Apr 2025 2:30 PM
The Bridge, Dumfries, DG2 9AW
Description
Summary
Welcome back to the exciting world of artificial intelligence! Today, we'll explore how computer vision and deep learning join forces to enable machines to "see" and interpret the visual world around them.
Eligibility
You need to be a Dumfries and Galloway region resident at the age of 19 and above, the caveat to that is that you're a school leaver with no positive destination in mind.
Pre-Requisite
This course is designed for programmers with a basic understanding of deep learning concepts and experience with Python libraries like NumPy and Pandas. Prior knowledge of computer vision fundamentals (image processing, image classification) would be beneficial.
Course Duration
Two Days (1000 to 1430 hours GMT) with a 30-minute break in both sessions.
Course Objectives
- Gain a deeper understanding of computer vision tasks and challenges.
- Explore Convolutional Neural Networks (CNNs) for image recognition and object detection.
- Learn advanced CNN architectures for computer vision applications.
- Implement deep learning models for image classification, object detection, and image segmentation.
- Understand techniques for data augmentation and improving model performance.
Outline
Day 1
Introduction to Computer Vision (1 hour)
- Overview of computer vision tasks (image classification, object detection, segmentation)
- Image processing techniques for pre-processing and feature extraction
- Applications of computer vision in various domains (autonomous vehicles, robotics, medical imaging)
Convolutional Neural Networks for Vision (1.5 hours)
- Review of CNN architecture (convolutional layers, pooling layers)
- Understanding filters, activation functions, and backpropagation in CNNs
- Pre-trained CNN models (VGG16, ResNet) and transfer learning
Hands-on Lab 1: Image Classification with CNNs (1 hour)
- Implementing a CNN model from scratch for image classification using TensorFlow or PyTorch
- Training the model on a computer vision dataset (e.g., CIFAR-10)
- Visualizing feature maps and interpreting model behavior
Day 2
Advanced CNN Architectures for Vision (1.5 hours)
- Exploring deeper and more complex CNN architectures (Inception, ResNet)
- Understanding residual connections and their benefits
- Object detection using CNNs (YOLO, R-CNN)
- Convolutional layers for image segmentation (U-Net)
Hands-on Lab 2: Object Detection with CNNs (1.5 hours)
- Implementing an object detection model using a pre-trained CNN (e.g., YOLOv5)
- Training the model on a custom object detection dataset
- Visualizing bounding boxes and evaluating object detection performance
Data Augmentation and Model Optimization (1 hour)
- Techniques for data augmentation to improve model generalization (random cropping, flipping)
- Regularization techniques (dropout, L1/L2) to prevent overfitting
- Hyperparameter tuning and best practices for training computer vision models
Course Wrap-up (30 min)
- Q&A and Discussion on computer vision using deep learning
- Introduction to advanced computer vision topics (e.g., image generation, image style transfer)
- Resources for further learning and exploring state-of-the-art computer vision models
- Course Feedback
Note: This course emphasizes hands-on learning through labs that introduce participants to building deep learning models for computer vision tasks. The course focuses on practical implementation and techniques for improving model performance within the one-day timeframe. Pre-trained models and datasets will be utilized to expedite the learning process.
Location
The Bridge, Dumfries, DG2 9AW