one of the finest videos on yolo available on internet. contains intuitive as well as detailed explanation (right from research paper). concepts like these are hard to explain in so much detail. thanks a lot for the amazing work, cheers!
this is an absolutely excellent explanation of YOLO algorithm. I have been going bonkers looking for a proper explanation and there is nothing like this one. Thank you so much for working so hard on this.
@@prathameshdinkar2966 Yeah, I started with Canva, thought will get to Manim if needed, but then canva fulfilled all my requirements(as of now), so never got to it.
Do you have any plans to make videos on detail mathematical explanation of reinforcement learning and its implementation? Will be highly interested to know about it.
Hello, Yes I intend to do that but its going to take some time to get there. Primarily because I would myself need to have a clear and in-depth understanding of it, which I dont as of now. But RL is definitely something thats on my list
Yes, I plan to start those once I have covered a decent chunk of detection papers. Then in parallel will also release videos on segmentation and later tracking.
Hello, Yes definitely, thats exactly whats in plan for this detection series. I am planning to do SSD first and then yolov2-3, yolov5 and so on. Ultimately its going to cover all yolo versions.
Let's say there is an image with 2 concentric spheres, hence both object belong to the same anchor box. Will YOLO v1 be able to detect both? Asking in reference to @7:46
Assuming both spheres would belong to the same class, and with the value of B(number of bounding boxes predicted per grid cell)=2, then yes YOLO v1 should be able to predict both.
@13:00 , The net loss function = Localisation loss + Confidence loss + Classification loss. But Isn't the confidence loss simillar to localisation loss? Also, what are target confidence scores in the Confidence loss is it also (1/0)?
Localization just predicts the xywh coordinate. Confidence predicts the objectness(the target of it is 0 or 1). I talk about targets in more detail @22:07 So localization answers the question 'where the object is' whereas Confidence answers the question 'Is there even an object'.
one of the finest videos on yolo available on internet. contains intuitive as well as detailed explanation (right from research paper). concepts like these are hard to explain in so much detail. thanks a lot for the amazing work, cheers!
Thank you for this comment :)
There're a lot of explaining Yolo video on youtube but this is still a clearest explaining video i 've ever seen. Appreciate your great works.
Thank you for the kind words :)
thank you so much, this video is very helpful to me. you are very generous.
I'm happy that the video ended up being of help to you :)
this is an absolutely excellent explanation of YOLO algorithm. I have been going bonkers looking for a proper explanation and there is nothing like this one. Thank you so much for working so hard on this.
Thank you for the appreciation and your kind words :)
Nice tutorials.
Keep on create a videos such this we will support.
Thank you for your support :)
We want more videos like this ❤
great explanation
great explanation👍
*Github Implementation Link* - github.com/explainingai-code/Yolov1-PyTorch
Nicely explained! Can you tell how do you create the animation, is there any library?
Thank you! The animations are all done using Canva. Bulk of the video is created in canva and then I use imovie at the end for some editing.
@@Explaining-AI Ohk, nice I thought you were using Manim
@@prathameshdinkar2966 Yeah, I started with Canva, thought will get to Manim if needed, but then canva fulfilled all my requirements(as of now), so never got to it.
thanks alot
Do you have any plans to make videos on detail mathematical explanation of reinforcement learning and its implementation? Will be highly interested to know about it.
Hello, Yes I intend to do that but its going to take some time to get there. Primarily because I would myself need to have a clear and in-depth understanding of it, which I dont as of now. But RL is definitely something thats on my list
@@Explaining-AI glad to know that.
dear may you explain object segmentation and tracking too, please
Yes, I plan to start those once I have covered a decent chunk of detection papers. Then in parallel will also release videos on segmentation and later tracking.
Hi!
Could you make a video like this on the latest YOLO models? or maybe mention what they have done differently in the latest model implementations
Hello, Yes definitely, thats exactly whats in plan for this detection series. I am planning to do SSD first and then yolov2-3, yolov5 and so on. Ultimately its going to cover all yolo versions.
Let's say there is an image with 2 concentric spheres, hence both object belong to the same anchor box. Will YOLO v1 be able to detect both? Asking in reference to @7:46
Assuming both spheres would belong to the same class, and with the value of B(number of bounding boxes predicted per grid cell)=2, then yes YOLO v1 should be able to predict both.
@13:00 , The net loss function = Localisation loss + Confidence loss + Classification loss. But Isn't the confidence loss simillar to localisation loss? Also, what are target confidence scores in the Confidence loss is it also (1/0)?
Localization just predicts the xywh coordinate. Confidence predicts the objectness(the target of it is 0 or 1). I talk about targets in more detail @22:07
So localization answers the question 'where the object is' whereas Confidence answers the question 'Is there even an object'.