YOLO (You Only Look Once) algorithm for Object Detection Explained!

In this video, I've explained about the YOLO (You Only Look Once) algorithm which is used in object detection.

Object detection is a critical capability of autonomous vehicle technology. It’s an area of computer vision that’s exploding and working so much better than just a few years ago.

YOLO is a clever convolutional neural network (CNN) for doing object detection in real-time. The algorithm applies a single neural network to the full image, and then divides the image into regions and predicts bounding boxes and probabilities for each region. These bounding boxes are weighted by the predicted probabilities.

YOLO is popular because it achieves high accuracy while also being able to run in real-time. The algorithm “only looks once” at the image in the sense that it requires only one forward propagation pass through the neural network to make predictions.

After non-max suppression (which makes sure the object detection algorithm only detects each object once), it then outputs recognized objects together with the bounding boxes.
With YOLO, a single CNN simultaneously predicts multiple bounding boxes and class probabilities for those boxes. YOLO trains on full images and directly optimizes detection performance. This model has a number of benefits over other object detection methods.

Some research papers on YOLO for better understanding of the algorithm:

#yolo #ObjectDetection #CNN #Python

[sagnikroy]

Thanks, sir. Your content helped a lot. Everybody just codes and moves on, but nobody tells how it happens. Thank You

[totally_insane]

Very lucid explanation and easy to understand. Learned a lot from this video alone, thanks and keep it coming

[mohammadyahya]

The best explanation on YOLO so far. Thank you.

[rashibhardwaj]

Amazing video. Thank you for explaining everything in just one video😃

[conOC]

Simple, clear and instructible. Perfect to introduce to YOLO. SO GOOD

[jaysoni]

Amazing explanation with enough time thanks for saving my time

[priyathirumalainambi]

Thanks balaji. You taught really well. Pls upload more videos. will be more useful

[venupunna]

really good simplification of yolo part1 Thankyou

[bethineedilakshmideepak]

@Balaji Srinivasan, Sir you explained exactly like Andrew ng in a detailed manner. Happy to come to know about your channel

[srinivasand]

Thanks much balaji. This will help me in my project preparations!

[mazharmumbaiwala]

any resources to the newer or better methods to solve the limitations of anchor boxes?
what if my image has 100 instances of different objects to be detected, can someone point a link or mention them

[jeffreyeiyike]

please i want to know which tensor or vector of the images saved. all I see is the bounding box and classification and probability

[apurbaroy]

Is it possible to integrate the YOLO algorithm with arduino or raspberry pi using a webcam?

[roobanrajr]

Bro today yenaku interview coding test iruku ....object detection model built pana solirukanga help pana mudiyum ma ?I have one two day to complete the code

[wcottee]

I missed something...for training and testing we have images plus bounding boxes in our inputs. But the final model input is image only. How is this handled?

[mehulmak]

If y output only detect one object at a time then how come we can have multiple object detected in single frame at a time?

[jayeshbagul]

Excelent it really benifical for me Thank you for your guidance

[praveenreddy]

I have doubt could you please clear this...Suppose consider 3 X3 Grid (grid1, 2, 3, 4, 5, 6) and consider a image ie car is spread over 2 grids (5th and 6th grids ) For Grid 5th, Yolo through CNN operation identifies image and its bounding box and vector cordinates are predicted covering two (5th and 6th) grid cells . Now for 6th grid also same operation will be applied . So now after whole grids operation does.5th and 6th grid predictions combined through NMS and IOU to single prediction where image is exactly PRESENT ? Is my understanding correct?

[punithpuni]

Sir I have a doubt please help me, you told that:
1) Output layer consists of both classification(pc, c1, c2, ...) and bounding box values(bx, by, bh, bw) i.e, its a regression.
2) At 2:45 you told that for ouput layer softmax activation is applied, but how can a softmax activation be applied on bounding box values which is regression.
3) Ok let me assume that as the width and height values of Image and grid will be between 0 and 1 their may be a chance of using softmax, because softmax activation output will be between 0 and 1, but Iam not sure about this. But at 17:05 you told that in some cases in output layer bounding box width and height can be more than 1, but softmax which is applied to output layer can give values between 0 and 1, then how can bounding box width and height get the value more than 1.
4) Softmax when used in output layer it will consider bounding box values also as classes, so how can softmax be used in output layer.
Can you please solve my confusion.

[DeepakSaini-sgpq]

Great explanation thank you 😊
#Subscribed