Automotive Camera [Computer Vision, Deep Learning] - 2A

Python-based algorithm development, real camera data from ADAS vehicle, image processing, object detectors, UML

What you'll learn

Understand real data of front camera collected from ADAS vehicle

Learn and implement step by step camera image processing and object detection modules in python 3.x

Develop software modules using object oriented programming and UML digrams in python

Acquire necessary practical skills of camera-based perception algorithm development necessary for ADAS / AD industry

Requirements

Working computer with internet facility

Basic mathematics - matrix, vectors, probability, transformations, etc.

Good Python 3.x knowledge of object-oriented programming concepts like classes, objects, inheritance, etc.

Basics of OpenCV, Numpy, and Pytorch packages in Python.

Able to configure and use an IDE like Visual Studio code, pycharm, or other.

Able to install various Python libraries using pip, conda, or others.

Description

Perception of the Environment is a crucial step in the development of ADAS (Advanced Driver Assistance Systems) and Autonomous Driving. The main sensors that are widely accepted and used include Radar, Camera, LiDAR, and Ultrasonic.This course focuses on Cameras. Specifically, with the advancement of deep learning and computer vision, the algorithm development approach in the field of cameras has drastically changed in the last few years.Many new students and people from other fields want to learn about this technology as it provides a great scope of development and job market. Many courses are also available to teach some topics of this development, but they are in parts and pieces, intended to teach only the individual concept.In such a situation, even if someone understands how a specific concept works, the person finds it difficult to properly put in the form of a software module and also to be able to develop complete software from start to end which is demanded in most of the companies.This series which contains 3 courses -  is designed systematically, so that by the end of the series, you will be ready to develop any perception-based complete end-to-end software application without hesitation and with confidence.Course 1 (already published and available online) - focuses on theoretical foundations Course 2A  (This course) - focuses on the step-by-step implementation of camera processing module and object detector modules using Python 3.x and object-oriented programming. course 2B (to be published very soon) -  focuses on the step-by-step implementation of camera-based multi-object tracking (including Track object data structures, Kalman filters, tracker, data association, etc.) using Python 3.x and object-oriented programming. Course 2A  - teaches you the following content (This course)In the complete course, you will develop a camera perception pipeline with 20+ classes using object-oriented programming in Python 3.x.You will implement a step-by-step camera image processing software module in Python 3.x to load real camera data collected from ADAS vehicles and preprocess it for the object detection module.You will develop a step-by-step complete object detection module in Python to detect various road users using FasterRCNN, SSD, YOLOv5, and YOLOv8.Learn systematically how to develop camera-based software for jobs and projects.[ATTENTION]:This is an advanced course in this field, so please fulfill all the stated prerequisites before you start this course. [Disclaimer]:Algorithms developed throughout this course are only for learning purposes. Learners must not use them directly in their projects or work without enough tests, debugging, and modifications according to requirements.[Suggestion]: Those who want to learn and understand concepts can take course 1 only. Those who want to learn and understand concepts and also want to know and do programming of those concepts should take all three courses 1, course 2A, and course 2B. 

Overview

Section 1: Introduction

Lecture 1 Why this course?

Lecture 2 Pre-requisites

Section 2: Image Preprocessing with python - Part 1

Lecture 3 Camera perception pipeline - a quick overview

Lecture 4 Understand and download camera data

Lecture 5 Image Preprocessing module - concept

Lecture 6 CameraPosition class - concept

Lecture 7 CameraPosition class - code

Lecture 8 TimeStamp class - concept

Lecture 9 Timestamp class - code 1

Lecture 10 Timestamp class - code 2

Section 3: Image Preprocessing with python - Part 2

Lecture 11 Image class - concept

Lecture 12 Image class - code 1

Lecture 13 Image class - code 2

Lecture 14 ImageReader class - concept

Lecture 15 ImageReader class - code 1

Lecture 16 ImageReader class - code 2

Lecture 17 Image Visualizer - code

Lecture 18 Complete Image preprocessing module - testing code

Lecture 19 Coding Activities - 1

Section 4: Object detection data structures with Python

Lecture 20 Object category and 2D bounding boxes - concept

Lecture 21 Object category enum class - code

Lecture 22 ImageObject2D class - code

Lecture 23 ImageObject2DList class - code 1

Lecture 24 ImageObject2DList class - code 2

Lecture 25 Different ways to represent 2D bounding boxes - concept

Lecture 26 Transform bounding boxes in different forms - code

Lecture 27 Coding Activities - 2

Section 5: Object detector software module with python - part 1

Lecture 28 Object detector software module - concept with UML class diagram

Lecture 29 Object Detector Abstract class - concept

Lecture 30 Object Detector Abstract class - code

Lecture 31 FasterRCNN object detector class - concept

Lecture 32 FasterRCNN object detector class - code 1

Lecture 33 FasterRCNN object detector class - code 2

Lecture 34 ObjectDetectorType enum class - concept

Lecture 35 ObjectDetectorType enum class - code

Lecture 36 ObjectDetector class - concept

Lecture 37 ObjectDetector class - code

Lecture 38 Test complete pipeline with Faster RCNN - code

Lecture 39 Visualizer class method for object detection - code

Section 6: Object detector software module with python - part 2

Lecture 40 FileWriter class and FileType enum class - concept

Lecture 41 FileType enum class - code

Lecture 42 FileWriter class - code

Lecture 43 SSD with VGG 16 Object Detector class - concept

Lecture 44 SSD with VGG 16 Object Detector class - code

Lecture 45 YOLOv5ModelSize and YOLOv5 class - concept

Lecture 46 YOLOv5ModelSize enum class - code

Lecture 47 YOLOv5 class - code 1

Lecture 48 YOLOv5 class - code 2

Section 7: Object detector software module with python - part 3

Lecture 49 YOLOv8ModelSize and YOLOv8 class - concept

Lecture 50 YOLOv8ModelSize enum class - code

Lecture 51 YOLOv8 class - code

Lecture 52 CameraPerception class- concept

Lecture 53 CameraPerception class - code 1

Lecture 54 CameraPerception class - code 2

Lecture 55 Test complete pipeline with main.py - code

Lecture 56 Revisit complete UML class diagram

Lecture 57 Coding Activities - 3

Section 8: Wrap up

Lecture 58 Metrics used for validating object detection performance

Lecture 59 Congratulations…

Anyone interested in camera-based perception algorithm development for ADAS / AD using python,Students, researchers, hobby people, etc. who wants to learn