Computer Vision by using C++ and OpenCV with GPU support

Name: Computer Vision by using C++ and OpenCV with GPU support
Rating: 3.8 (31 reviews)

Learn how to Use OpenCV with GPU support on Ubuntu OS

Created byIbrahim Delibasoglu

Last updated 9/2022

English

English [Auto],

What you'll learn

Computer vision
OpenCV
Computer vision app on Ubuntu OS
C++ skills

Course content

7 sections • 27 lectures • 3h 22m total length

Course Overview3:38
Set up the environment, compile open source code, and explore computer vision with C++ and OpenCV on CPU and GPU, including color space transformations, filtering, trackbars, and background subtraction.

Driver installation5:46
In this video, you will see the different ways to install Nvidia driver on Ubuntu 18.04. If you wanna how i installed the driver directly in easier way, you can skip 05:20 :)
Cuda toolkit installation1:14
Compile OpenCV from source with CUDA support part-16:18
Compile OpenCV from source with CUDA support part-25:07
Note: install cuda toolkit 10.2 for Nvidia optical flow support

error fix while compiling opencv with cudaoptflow
error: no such file nvOpticalFlowCommon.h
Copy header files from OpticalFlowSDK/NvOFInterface/ to opencv_contrib/modules/cudaoptflow/src
Python environment for flownet2-pytorch9:27
This is to be able run deep learning based dense optical flow implementation flownet 2.0

Read camera & files in a folder (C++)10:54
Learn how to read images from a folder, read videos, and capture webcam frames using C++ with OpenCV, including listing, filtering JPEG files, sorting filenames, and displaying frames.
Edge detection (C++)7:34
Explore edge detection in images using the canny method in OpenCV with C++, converting to grayscale, applying hysteresis thresholding with two thresholds, and noting GPU acceleration.
Color transformations (C++)7:13
Explore HSV color transformation in OpenCV with C++, converting images to HSV, applying color filtering with thresholds, and extracting white lines and colored regions via mask images.
Using a trackbar (C++)5:59
Image filtering with CUDA (Introduction to using OpenCV GPU methods on C++)13:13

Background segmentation with MOG (C++)3:58
Learn background segmentation with MOG in C++ using OpenCV, subtract reference and current frames to produce a foreground mask, and explore OpenCV background subtraction methods and implementations.
MOG and MOG2 cuda implementation (C++ - cuda)3:25
Explore CUDA-accelerated MOG and MOG2 background subtraction in C++ with OpenCV, transferring frames and masks to the GPU, implementing background segmentation, and evaluating GPU performance against CPU.
Special app: Track class5:57
Special app: Track bgseg Foreground objects8:01
Extend the background subtraction workflow by applying gaussian filtering and morphological closing to refine the foreground mask, compute contours, filter regions, and track objects with a tracker.

A simple application to prepare dataset for object detection (C++)7:39
Show how to build a simple dataset for object detection with OpenCV in C++, capturing image patches via a webcam rectangle and saving them for model training.
Train model with openCV ML module (C++ and CUDA)12:35
Object detection with openCV ML module (C++ CUDA)6:23
Explore a basic object detection workflow in c++ with opencv's ml module using svm and histogram of gradients, including dataset creation, feature extraction, and webcam testing on gpu.

Optical flow with Farneback (C++)8:11
Demonstrate optical flow concepts with C++ and OpenCV, covering sparse (Lucas-Kanade) and dense (Farneback) methods, visualizing results and saving flow videos, with a deep learning alternative.
Optical flow with Farneback (C++ CUDA)5:50
Optical flow with Nvidia optical flow SDK (C++ CUDA)5:04
Optical flow with Nvidia Flownet2 (Python)5:06
Apply optical flow with Nvidia FlowNet2 in Python by running a script that processes consecutive image frames and outputs flow results.
Performance Comparison6:29
Compare the results of multiple optical flow methods across images and videos, assess their performance at varying resolutions, and note real-time feasibility and visualization differences in Python.

Moving Camera Motion Estimation12:03
Warping via CUDA implementation6:50
Foreground object detection with feature comparison14:34
You will learn how to eliminate wrong detections (False Positives) in foreground detection for background movements.
Custom object detection with CNN based features13:58
You will see an example for Pytorch C++ API. We will save a model for feature extraction in the Python, and then use that model in C++.

By using extracted features, we will classify two target objects.

Requirements

Must have basic programming knowledge

Description

In this course, you are going to learn how to install Nvidia driver on ubuntu OS and compile OpenCV with GPU support. And, you will see how to use opencv GPU functions to accelarate your applications. Also you are going to learn how to setup nvidia flownet2-pytorch environment in python.

*** Watch the Introduction video for more details! ***

If you firstly follow my other course "Learn Computer Vision with OpenCV and Python", you will learn more beginning level information in computer vision, and then it will be better for you to see different examples with C++ and GPU enabled functions in this course.

Who this course is for:

Software developers
Students want to work on Computer Vision

Computer Vision by using C++ and OpenCV with GPU support

What you'll learn

Explore related topics

Course content

Introduction1 lecture • 4min

Set up Necesssary Environments5 lectures • 28min

Introduction with a few basic examples!5 lectures • 45min

Background segmentation4 lectures • 21min

Object detection with openCV ML module (C++ CUDA)3 lectures • 27min

Optical Flow5 lectures • 31min

Bonus/Different Examples4 lectures • 47min

Requirements

Description

Who this course is for: