วีดีโอ

for tx ty they are relative to grid cell why we are not multiplying with 64 like in yolov1

This implementation has overflow-error with very small or large values for x

bro start making videos once again, your explanataions are very good

Thanks a lot for this video, this was really helpful

Thanks a lot for your video, this helped me a lot to understand its working

Also, how is it possible that beta/gamma make a non-symmetric, non-normal distribution like in your picture? It can only give a different normal distribution that may not have mean 0 and std 1. But it's a linear transformation of a std normal, so it must still be normally distributed. Right?

I'm a little confused...you say at around 10:21 that in each mini batch we normalize the x's, which usually denotes the inputs and this doesn't seem like what I think of as batch normalization. But then you go on to say in the visual explanation afterwards what I expected and what I think is correct which is that we normalize the Zs in layers after the input layer before sending them to the activations. So, to be be clear: we don't normalize the batch inputs themselves, just the Zs, right? Or is it both? By the way, this is excellent material--it's just that this one point was confusing me a little

woah, i am happy i found this!

This is amazing could you do a transformer series!

Thanks

Glad! i found this legendary explanation

Best

Bro ❤ that's so better than 200 usdt lectures

hiii will u post yolo v5 ? this is the best playlist

thank you! clear and direct.

Thank you very much for the best explaination of yolo papers on youtube. I have a question of loss calculation on multiscale training. This affects the number of output grid (WxHXS) used in loss calculation when input image(WxH) size changes. How does the loss calculation maintain consistency for this training scheme?

could you please mention source of the mathematical explanations it would be great help.

The best explanations of YOLO on TH-cam. Period. Thank you!🙏

really great job sir! waiting for more lately versions

You rock!!! It was very detailed. Clearly, you have out a lot of work into this. Thank you so much🙏🙏🙏🙏🙏🙏

Just wow!

The best explanation for this concept, kindly keep making such content. Thanks a lot.

thanks for giving us perfect video, love you bro

thanks for helping us a lot in learning, truly appreciate your work

very detailed explanation, Thanks for making it more clear. I believe i didn't find any such video with the way you explained the things in deep. I have a doubt when you said total loss = obj loss+no obj loss, In the example you considered only 2 grid cells has an object which means obj loss is calculated for those 2 grid cells and remaining 47 grid cells falls under no obj loss right?

Underrated. Keep going man!

thank you i could have nerver learn and understand any better then your video

i still want to know how to code this stuff up? theory can be found everywhere but how to dive into doing it ourselves?

this is such a great vid

great explanation! thanks!

Very cool project! I have only one question: based on the video, you achieve 71.6% accuracy for 10 epochs, but I can't go above 52-54%. However, you also only have 56.94% accuracy in your github project. What is the reason of this? I'm very new to DL/MI, that's why I'm asking. I tried so much that I went up to 20 epochs, but the end result was almost exactly the same as that of 10 epochs. However, the accuracy increased by half per epoch... Thank you very much for any suggestions!

Just like sir said the higher level feature map or smalle feature map are mainly for bigger object and lower level feature map or bigger feature map are for smaller object,due to this various scale are neded for differenet feature map , can anyone tell me in simple words what is 'scale' here

One of the best I have come across so far, Could you please continue the series.

Great work on the explanation, very well summarized!

nice video maboy, the best theorical explanation of why numpy is that good

great video, now I finally understand it :) could you just please clarify why in 22:32 only 2 grid cells contain objects? the woman appears in a few other cells as well, so why only two?

IM EATING THIS UP THANK YOU

Hello Sir First upon thanks for your all videos, Sir i request you please upload Video on YOLOv8 architecture and How to improve accuracy in yolov8 model.

Fantastic explaination. Super clear

underrated af

WOW what a clear and simple explanation

please make video soon on remaining yolo variants

thanks!

what a great material - so well structured. thx for your efforts - beautiful!!!

Dear Author, thx for your great work! you are awesome!

Best YoloV3 video!

Excellent Explanation . Will not forget concept now till going to grave

Hi! thank you for your wonderfull explanation! Unfortunately in the original paper there are many unclear moments. Your video helped me a lot. But i still have some questions. 1) "Grid cell is "responsible" if the center of bbox falls into it." In training data we have annotated bboxes. But in test data there are no annotated bboxes and therefore centers. So which grid cell will be "responsible" in that case? 2) if c < threshold, then we simply nullify all the values in the vector or we should train the model to nullify the vector on its own? 3) if only 2 grid cells (in your case) predict the coordinates of bboxes, what is the use of the other 47 grid cells (are the useless at all or not?) 4) How one small grid cell (64x64) predicts a box for an object that is a way bigger than this cell (450x450)? 5) Why you are telling that there are only 2 object cells, if the woman overlap at least 6 cells? Maybe you mean only 2 "responsible" cells?

And, I have a question for the "Not 0 centered function" section : What if you use the activation function only up to the last hidden layer. You just don't put any to get the final result.. The weights between the hidden layer and the output might be negative and the backpropagation won't be changing the sign of it.. will it ?

Yes it's aaawesome !! I was searching a beginner video for weeks but they are all only "beginner friendly" by name.. It comes really fast into complicated maths subjects and you don't get to understand the compution process behing it.. You explain everything :) it's really cool ! TY