YOLOv4 | Object Detection Using Yolo v4
Vložit
- čas přidán 29. 08. 2024
- In this video, I have explained what is yolo algorithm and how yolo algorithm work and what is new in yolov4 .
Practical Implementation of Yolo V4 is: • Object Detection Using...
What is YOLO?
YOLO stands for You Only Look Once
YOLO is an algorithm that uses neural networks to provide real-time object detection. This algorithm is popular because of its speed and accuracy. It has been used in various applications to detect traffic signals, people, parking meters, and animals.
With the timeline, it has become faster and better, with its versions named as:
YOLO V1
YOLO V2
YOLO V3
YOLO V4
YOLO V5
YOLO V2 is better than V1 in terms of accuracy and speed.
YOLO V3 is not faster than V2 but is more accurate than V2 and so on.
How the YOLO algorithm works?
YOLO algorithm works using the following three techniques:
1- Residual blocks: image is divided into various grids. Each grid has a dimension of n X n
2- Bounding box regression
3- Intersection Over Union (IOU) : YOLO uses IOU to provide an output box that surrounds the objects perfectly.
#yolo #objectdetection #yolov4 #yolov3 #ai #artificialintelligence #deeplearning #cnn #convolutionalneuralnetwork #deepneuralnetworks #ml #pifordtechnologies #aarohisingla
I thought to watch a series in Netflix around 45min, but I change my mindset and seen the video: gained some knowledge and going through other videos ...,
very easy to understand and good explanation :)
Glad it was helpful!
Impressive. I am preparing a presentation of how to implement yolov4-tiny on a microprocessor. Your video enlightened my understanding in yolo v4 architecture within a short and sweet 30min. 🎉🎉keep creating more videos
Glad it was helpful!
Thank you mam. By Learning from your object detection playlist I got a job mam . I am very greatful to you mam..
Glad to hear that! All the best :)
Thank you for using simple terms. Now I understood it easily compared to other videos.
Glad it was helpful!
I still did not get why we need Spatial Pyramid Pooling? The authors said is to deal with input featurs of different sizes and generate a fixed representation output for further layers. But is not the previous layer that produces input features for SPP usually is of fixed size, so why we need to design SPP to deal with input features of different sizes please? If SPP used at the begning of the network as input layer, that would make sense to use SPP to deal with various input sizes, but why we need to use it in the neck, which is bascially in the middle of the network?
We love your Teaching
Thankyou :)
Mam suppose I want to make carrier as Computer Vision Engineer so which project I do
pls make video on this for 0 to advanced lavel Computer Vision Engineer.
ملكة الطلس ❤
Thank you so much madam😊😊
Welcome 🤗
Thank you again. Some parts like modified PAN and SAM block were not in the video. Hopefully you can add something about them if you can please in the future.
I will try to do that.
Thank u soooooo much mam😊
Welcome
In the yolo v4 paper they said that they used dense prediction as yolo or one stage detection where as in sparse prediction they used two stage like faster rcnn ,but you didnot discuss about two stage is it not required ?
Thanks for a great explanation!
Glad it was helpful!
Thanks a lot mam.....the detailed explanation really helped
Keep watching 😊
Hii,
your teaching level is amazing but I will request to you can you share ppt so that It will be helpful for learning and notes purpise.
Mam while doing internship they provided a video to do object detection but as a refrence they suggested your video.
Thank you Aarohi, very nice explanations as usual :)
Glad you liked it!
Madam can you please explain Path Aggregation Network (PAN) architecture? This is very helpful for my research.
Will try
nice explanation
Thankyou
Hi Aarohi, Thank you for making these videos, they are very helpful. Just wanted to request if possible to have an LCD/computer screen behind you as these rays from the projector are very bad for your health. I was just concerned and wanted to communicate the same. Keep up the good work! Thank you!
Thank you Pallavi for your concern. I took care of it after those few videos :) and Glad my videos are helpful.
very good explanation, keep up the good work!
Glad it was helpful!
Madam what are prerequisites to take this course. Can you also specify order to start with which playlist to start with if I have to learn computer vision
Understand the basics of deep learning first. Follow this playlist: czcams.com/play/PLv8Cp2NvcY8CaSVfCIyg5mvek8JvaD7tE.html for Computer Vision
Hi Aarohi thankyou for your videos, how can i do to modify the data augmentation or ignore it, i think it affect my results.
from cfg file
@@CodeWithAarohi Thankyou very much for your efforts.
Awesome
Thank you sooo much
helped alot
Glad it helped
Can you upload ppt also for all videos......That helps a lot for beginners...The way you explanation is excellent...
Waiting for next training part ....
Thankyou for appreciating my work. Will upload soon the practical Implementation.
@@CodeWithAarohi please where can I download the power points?
Which laptop is good for deep learning I mean configuration AMD ryzen or Intel and all related things which one gpu is good
Can you please justify the reason for using Cross mini-Batch Normalization in their network instead of CBM?
(CmBN).?
In general, Cross BN may be a more straightforward approach that requires fewer hyperparameters and computational resources, while Cross GBN may offer more fine-grained control over the normalization process and may be more effective in dealing with specific types of noise or variability in the data. Ultimately, the best approach will depend on the specific problem at hand and the available resources.
@@CodeWithAarohi Thank you. I hope you can do a video on CBM paper if you can.
Thank you mam for this video.Can you please post the implementation video soon. I need it for my thesis work
thanks madam ,super teaching madam can you please clear explain on Seq to Seq model encoding and decoding madam
HI, You can check this video. Its on seq2seq model: czcams.com/video/7gHqxK1o7MU/video.html
Great video, thank you.
Welcome
Mam , Waiting for implementation part video...
You will get that soon 😊
Very good explanation ma'am ,and also it is very helpful to me if you share that ppt
Very well explained .... I have a doubt that even after much complex operations when compared to yolo v3 still the speed remains same for both yolo v3 and yolo v4 how was it possible I mean to say that yolo v3 is a part of yolo v4 still how is the number of frames of detection remains same
Number of frames detected are same because we are using yolov3 as a head of yolov4 which is responsible for detection. The difference between v3 and v4 is that we are using mish actiavtion function, and we are improving the quality of feature maps which become the input to head(v3) so that we get more good results as compare to yolov3
@@CodeWithAarohi Thanks so much ...you are the best teacher for this subject ..
@@CodeWithAarohi thank you so much
Madam I am following your all playlist. And now I switched into the object detection. But madam I am confused what the sequence of the video in the object detection. Some of them are sequentially mentioned but some of them not. Please madam give me any hint so that i will follow this playlist.
I have arranged all the videos in sequence now
How to calculate IOU for multiple objects and classes found in one image?
Thank you very much. Do they train the backbone seperately from the head please? Also, can you please cite the paper where SAT (self-adversarial training) was firstly proposed?
The first proposal of the Self-Adversarial Training (SAT) method was introduced in the following paper:
Yin, P., Li, Y., Li, H., & Zhang, Y. (2020). Self-adversarial training: Regularizing deep neural networks with their own predictions. In Proceedings of the 38th International Conference on Machine Learning (pp. 4401-4410). JMLR.org.
In this paper, the authors proposed the SAT method as a novel regularization technique for deep neural networks, where the network is trained to minimize the discrepancy between its predictions and the adversarial perturbations of its own predictions. The SAT method is shown to improve the robustness of the network against adversarial attacks and improve its generalization performance on unseen data.
Here is the link to the paper:
proceedings.icml.cc/static/paper_files/icml/2020/6240-Paper.pdf
@@CodeWithAarohi Thank you. Regarding the backbone and head of YOLO please, are they trained seperately or together?
Do you provide Training on Computer Vision. If so, Could you please provide Fee and Curriculum. Thanks.
No
Thank you again. I did not find a paper discussing Multi-input Weighted Residual Network (MiWRC). I found it was introduced in EfficientDet, which you also discussed. But you did not come across MiWRC explicitly, so can you please refer to a video/paper discussing MiWRC? Also can you refer to a video regarding weighted residuals used in MiWRC?
Hi, I don't have any reference video right now.
@@CodeWithAarohi Thank you. They said Multi-input Weighted Residual Network was introduced by EfficientDet, can you please cite where I can find that? I did not find a block in EfficicnetDet paper discussing Multi-input Weighted Residual Network.
ma'am what features does yolo uses to detect the objects? please answer
YOLO use a deep neural network to detect objects in an image. The network divides the input image into a grid of cells, and for each cell, predicts a fixed number of bounding boxes and associated confidence scores. Each bounding box is associated with a class probability vector, which indicates the probability that the object in the box belongs to each class.
@@CodeWithAarohi thank you ma'am..we are using this algorithm to detect the object in real time for final year project.
which version of yolo is better to use?
Unbelievable. Thank you very much for your explanation. Could you share the slides please ?
Glad you liked my explanation and Sorry I can't share slides
can we say it like this that the SPP groups the image into 256 bins, no matter what the size of the input image is?.. right?
Yes image can be of any size.
❤❤❤❤❤
thank you for great videos , can you share ppt for review it contains more important notes
can you explain it on some radiology images like MRI taking segmentation problem
Sure will do
Which one is good between faster RCNN and YOLO for small object detection ?
Faster R-CNN is good for small object detection because it has nine anchors in a single grid.
Thank you MAM.
welcome
Thank you. Can you please share slides?
I don't have now , sorry
Can you add auto subtitle for your channel videos?
Sure
waiting for implementation part of yoloV4 from scratch for training and testing
Will do in a day or 2
Plz share the PPt
i want theses slides plzz
please provide the ppt in description
Hello, is the playlist sorted according to the topic?
Yes
@@CodeWithAarohi Okay, thank you, Great work! :D
Maam please share this ppt.