Train YOLO for Object Detection with Custom Data
What you'll learn
- Apply already trained YOLO v3-v4 for Object Detection on image, video and in real time with camera
- Label own dataset and structure files in YOLO format
- Train YOLO v3-v4 detector in Darknet framework
- Assemble custom dataset in YOLO format
- Convert existing dataset of Traffic Signs in YOLO format
- Build individual PyQt graphical user interface for Object Detection based on YOLO v3-v4 algorithm
Requirements
- Basic knowledge of Object Detection algorithms
- Basics on how YOLO works
- Intermediate knowledge of Python v3
- Basic knowledge of OpenCV
- Basics on how to work with Anaconda Environments
- Basics on how to work with PyCharm IDE or any other Python IDE
- Basics on how to work with Terminal Window or Anaconda Prompt
- To have Linux Ubuntu installed is optional, but recommended
Description
In this hands-on course, you'll train your own Object Detector using YOLO v3-v4 algorithms.
As for beginning, you’ll implement already trained YOLO v3-v4 on COCO dataset. You’ll detect objects on image, video and in real time by OpenCV deep learning library. The code templates you can integrate later in your own future projects and use them for your own trained YOLO detectors.
After that, you’ll label individual dataset as well as create custom one by extracting needed images from huge existing dataset.
Next, you’ll convert Traffic Signs dataset into YOLO format. Code templates for converting you can modify and apply for other datasets in your future work.
When datasets are ready, you’ll train and test YOLO v3-v4 detectors in Darknet framework.
As for Bonus part, you’ll build graphical user interface for Object Detection by YOLO and by the help of PyQt. This project you can represent as your results to your supervisor or to make a presentation in front of classmates or even mention it in your resume.
Content Organization. Each Section of the course contains:
Video Lectures
Coding Activities
Code Templates
Quizzes
Downloadable Instructions
Discussion Opportunities
Video Lectures of the course have SMART objectives:
S - specific (the lecture has specific objectives)
M - measurable (results are reasonable and can be quantified)
A - attainable (the lecture has clear steps to achieve the objectives)
R - result-oriented (results can be obtained by the end of the lecture)
T - time-oriented (results can be obtained within the visible time frame)
Who this course is for:
- Students who study Computer Vision and want to know how to use YOLO for Object Detection
- Students who know basics of Object Detection but want to know how to Train YOLO with New Data
- Students who study YOLO and want to Label Own Data in YOLO format
- Students who use already existing datasets for Object Detection but want to Convert them in YOLO format
- Young Researchers who study different Object Detection Algorithms and want to Train YOLO with Custom Data and Compare results with different approaches
Featured review
Instructor
I am PhD student in Intelligent Systems. Studying Computer Vision, Machine Learning, Image Processing. Developing algorithms for safety autonomous vehicles.
I have a BSc in Manufacturing automation where I obtained knowledge on how to improve production speed and quality by integrating more efficient equipment, like non-stop filtering, velocity and temperature control in real time, as well as optical sensors for sorting and classifying different types of products.
And I have an MSc in Intelligent Systems where I obtained extensive knowledge of machine learning, computer vision, and intelligent robotics. My final project was to develop Alarm-Warning system for mobile robot that has information about distances to the objects - Safe distance, Warning distance and Alarm distance. The system creates a kind of bubble around mobile robot with green, yellow and red zones preventing collisions with obstacles.
I have published research on using different dimensions of filters for convolutional neural networks (ConvNet) for effective classification of Traffic Signs. Trained ConvNet I deployed on the Web on Linux VPS and on the basis of Flask framework in order to have opportunity to test classification online.
Professional interests: Computer Vision, Convolutional Neural Networks, Autopilot Car's System, Autonomous Robots.