Computer Vision with OpenCV: HOG Feature Extraction

In this excerpt from "Autonomous Cars: Deep Learning and Computer Vision with Python, " Dr. Ryan Ahmed covers the Histogram of Gradients technique, and how OpenCV can use it to extract features from images using Python.

[КириллКлимушин]

It's amazing. So pleasant to see the person with a huge sparks in the eyes explaining such non intuitive concept. Deserves millions of likes

[feraudyh]

It's so pleasant to see someone who takes all aspects of the video so seriously as well as the enthusiasm and the explanation.

[thijsgelton]

This video is amazing. I need people to find this exact video. You explain it in a way that is easy understandable, unlike most of the other videos out there.

[princeneo]

This video is a just awesome, explained HOG in a very simple but detailed way

[tnmyk_]

How are the HOG features visualized at 10:45? What is being plotted exactly?

[arttartkhai]

This video is giving me a chance to survive my senior life. T_T

[shahnazfat]

sir, what is on the vertical axis of histogram . it is supposed to be no. as you said how many gradients having particular anle. but in the video it is shown as magnitude... kindly clarify

[jaiv]

Are HOG features currently the best method of image simplification?

[awaisahmad]

Amazing video

Amazing

I have covered all concepts regarding histogram of oriented gradient

[sureshpokharel]

Very good explanation on image gradients.

[tunguyenanh]

that's amazing video, your example help me alot to grasp this content

[premendrasrivastava]

Wow, loved it, very nice explanation👍

[shahnazfat]

nicey explained. you have gone to the depth. just clarify one my my doubt. at 8 15, how each pixel can have gradient magnitude and direction independently .. plz . as gradient is for moving from one pixel to other. with respect to other pixel we can find. not for each pixel right

[mohammadidreesbhatresearch]

I am confused about the different variants of it, for example, UoCTTI, what do we mean by directed and und-rected gradients.

[sahiltrivedi]

One would think step 2 lead to the loss of information. When you bin the angle values and their intensities you discard their positions. Images are highly spatial data hence ignoring the placement of the gradients and only vectoring their angles and magnitudes seems like a way to cut corners. It must negatively impact the accuracy of neural networks being trained on these images.

[akashrizvi]

I can't find next part of this video which is about code of it ???

[ishfaqzaman]

Each 8x8 cell is represented as a 1x9 HOG vector. How are all these vectors fed to the SVM? Do we add them up and feed the aggregate 9d vector or are the vectors appended? Thanks!

[haithamalnasrawi]

From where did you get 50, 70, 100, 120 ? >>>> Did you assume them?

[Integralsouls]

awesome video, love to see such well thought stuffs

[denismerigold]

How to create your own deep learning library?