YOLOv8 Comparison with Latest YOLO models 

Better doesn't mean good latency, every new model is better, faster and energy efficient from the previous one.

Depends on the model you trained as well. If you train a more narrow model, you tend to get better results overall.

I think if you can put up a detection fps then it will be better to point out the advantage of yolov8

@@andrespereira4852 YoloV8 is a huge leap. From a developer's standpoint it's significantly easier to train and deploy... well, besides some of the deployment/export functionality to other frameworks is still in the works. But having a CLI to interact with YOLO is sublime. It's the first one to have it.

@@ForTheOmnissiah i don't know, but as a beginner to object detection things, it feel weird to interact with CLI

Not really. The architectures are different. Yolov7 was built by the original devs (AlexeyAB and WongKinYiu) that did v4 and its derivatives. pjreddie (who originally developed the darknet framework) publicly said that Yolov4 is the defacto successor. Ultralytics (yolov5 & yolov8) uses a different framework written in Python. Darknet is written in C++.

Yolov8 is much faster than v7 in the last shot of the moving cars. v7 completely fails that task.

thanks

Was wondering the same thing

It seems that YOLOv5 detects more objects with greater accuracy than YOLOv8. However, YOLOv8 runs faster than YOLOv5. I will do a complete analysis of this.

@@pamudu123ranasinghe7 thanks!

@@pamudu123ranasinghe7 after the your analysis please share your opinion, I will have been waiting for that :)

It seems that YOLOv7 is better than those two. Just look how steady and accurate are its bounding boxes from 0:25 to the end of the video. And why are its boxes painted with so tiny borders?🤨 It looks like Ultralytics is like the old chinese version of products, "look, look, is better and fast, dont worry about if it doesn't have a paper, just buy it, please, take it"

huggingface.co/spaces/Pamudu/YOLO-Battlefield Try this on huggingface spaces. You can compare your desired YOLO model with other YOLO models to get an idea about real-time performance and detection accuracy.

Its an option, you can define the thickness of whatever model

Out of these models, YOLOv8 performs better on a CPU. It has a higher FPS rate, high detection accuracy and easy to use. However, there is a new addition to the family called YOLO-NAS performs even better. It has a much higher FPS rate and maintains good accuracy. I have added a comparison video on YOLOv8 and Yolo-Nas.Check it aslo ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-91p2SkSuZkc.html&ab_channel=Pamudu123Ranasinghe

@@pamudu123ranasinghe7 thanks bro

huggingface.co/spaces/Pamudu/YOLO-Battlefield Try this on huggingface spaces. You can compare your desired YOLO model with other YOLO models to get an idea about real-time performance and detection accuracy.

Yes you can use YOLOv8 for classification. Refer this: roboflow.com/model/yolov8-classification

YOLOv8 Refer to this for more optimization techniques for achieving a higher speed. github.com/pamudu123/object-detection-optimization

Try "Ultralytics HUB" app at appstore.They run YOLO models real time on phone. I am not sure yolo models can run locally on the phone. But you can host your model on a cloud platform and call it in through your flutter application.

In my experience, YOLOv8 is better. If you find it running slowly, you can speed it up by quantizing the model or converting it to a CPU-optimized version. Another option is to try the YOLOv8 nano model and reduce the input image size for faster inference. Check out the Ultralytics documentation for easy guidance.

drive.google.com/drive/folders/1XuLxUudlMVMnsUDR-kK9r_YXooT9yMGL?usp=sharing

yes sure, you can contact me via pamudu1111@gmail.com

YOLO medium models of each version is used for comparison

YOLOv8 is fast and easy to use in Windows 7 Refer to this for more optimization techniques for achieving a higher speed. github.com/pamudu123/object-detection-optimization

@@pamudu123ranasinghe7 thank u so much sir

You can use the following link to train a YOLOv8 model on a custom dataset: github.com/roboflow/notebooks/blob/main/notebooks/train-yolov8-object-detection-on-custom-dataset.ipynb universe.roboflow.com/wei-lun-wong-rlpuz/car-plate-detection-pnq5k You can use this dataset or another dataset from Roboflow Universe to train this. The dataset in this link only includes predefined classes, which may not cover all possible variations of license plates. For example, license plates can contain different numbers and letters, such as Roman numerals or characters from languages other than English. An alternative approach would be to train an object detector to identify the region of a license plate. After that, you can use an OCR method (easyOCR would be a good choice) to extract the characters from the license plate. This method will give more precise results.

Always an indian student that wants to copy code for college

What kind of anomalies are you planning to detect? If you use already trained YOLO models, they can only detect the 80 classes in the COCO dataset. tech.amikelive.com/node-718/what-object-categories-labels-are-in-coco-dataset/ If you need more clarification, please send more details about your project to my email.

@@pamudu123ranasinghe7 sir we are beginners we have no idea plz guide us , tell me some steps that we start! Thanks for replying sir ❤️

Yes sure, No problem at all

Kind of 😄. Altought its library ultralytics are one of the best for developers out there

It runs significantly faster than previous models. By quite a bit. That's a huge plus for many applications. In fact, most video-input applications require speed over accuracy.

huggingface.co/spaces/Pamudu/YOLO-Battlefield Try this on huggingface spaces. You can compare your desired YOLO model with other YOLO models to get an idea about real-time performance and detection accuracy.

huggingface.co/spaces/Pamudu/YOLO-Battlefield Try this on huggingface spaces. You can compare your desired YOLO model with other YOLO models to get an idea about real-time performance and detection accuracy.

yes sure, pamudu1111@gmail.com

Its the worst imo. Every time I train a yolov7 model, its a disappointment.

@@RandomGuy-df1oy I have used YOLOR model It gives me better results than YOLOv7

@@pamudu123ranasinghe7 What about YOLOv8? Currently, I still use v5 despite the new v6 and v7. I'm gonna try YOLOR too, thanks

@@pamudu123ranasinghe7 What about YOLOv8? I still use v5 despite the appearance of v6 and v7. I'm gonna try YOLOR too, thanks.

@@RandomGuy-df1oy But how about those custom datasets of yours? Are they good enough to get good accuracy?🤔🤨