Replies: 8 comments 41 replies
-
@Sukaina1 hello! 🚀 Customizing the YOLOv8 architecture, especially with modifications to layers or the backbone, can initially seem daunting, but let's break it down into manageable steps. For modifying layers or changing the backbone:
To load and train your custom model architecture in Python, you could use something like this: from ultralytics import YOLO
model = YOLO('path/to/custom_yolov8.yaml') # Load custom model architecture
model.train(data='path/to/your/dataset.yaml', epochs=100) # Train Tips:
Given your timeframe, prioritize learning the basics of neural networks and PyTorch (if using YOLOv8’s PyTorch implementation), as this knowledge will be invaluable in making these tweaks and understanding the underlying mechanics. Good luck with your thesis work! You're embarking on a challenging yet exciting journey. Keep experimenting and learning. 📚💡 |
Beta Was this translation helpful? Give feedback.
-
I have created a yaml file in yolov8 . i tried implementing gelan instead of cf2 . But to do that i need gelan.py or that is what the error says . so can u guide me on wt i should do . how i should proceed further ? |
Beta Was this translation helpful? Give feedback.
-
Thanks for replying @glenn-jocher @pderrenger. I am fairly new so I am unsure about the changes that should be done. I am trying to implement a small object detection with yolov8+sahi. and I have to make certain changes to the yolov8 architecture. combining cspnet+elan = gelan . |
Beta Was this translation helpful? Give feedback.
-
@glenn-jocher . Thanks for your explanation . There is no common.py - so I think you are mentioning about tasks.py - I have made changes in tasks.py,init , conv and blocks . but when I try to swap c2f layer with gelan - I am encountering the error of gelan not existing . So should I make changes at any other places. and from the error - there is a frozen error - can u explain on like - wt does each of the error entitle to ?Thanks for the helppp |
Beta Was this translation helpful? Give feedback.
-
thanks @glenn-jocher .so Load the best modelbest_model = torch.load('best.pt') |
Beta Was this translation helpful? Give feedback.
-
DID YOU GET ANSWER FOR YOUR ISSUE??? I Want to learn how to MODIFY THE ARCHITECTURE OF yolov8? CAN ANYONE HELP ME LEARN? ANY SOURCE WHERE TO LEARN THE MODIFICATION OF ARCHITECTURE OF YOLOv8 ??? |
Beta Was this translation helpful? Give feedback.
-
Hi @glenn-jocher , So to implement some additional feature to YOLOV8 . For example gradient blending technique to be added and new dataset need to be implemented. Need to modify Data loader or Data augment.py files need to be added. Please help on this . Regards |
Beta Was this translation helpful? Give feedback.
-
Hey, I am currently modifying the YOLOv8 Classification Model architecture by adding SE Blocks in it, I have also made some changes in engine/model.py in the train function to load the modified model instead of loading the default model as the training model. But now I am facing an error for an attribute being accessed which doesnt exist. I've mentioned it here in detail #17507 (comment) . Can anyone help me with this :') |
Beta Was this translation helpful? Give feedback.
-
I am an engineering student with very little know how about Deep Learning and Neural Networks. In my thesis work, I need to do some modifications in my trained yolov8 model(add/remove layers or weights, or change backbone). I have hardly two weeks for this.
Can anyone teach and guide me about this. I will be really thankful.
Beta Was this translation helpful? Give feedback.
All reactions