Open yxl23 opened 2 months ago
Thank you for your interest in enhancing YOLOv8 with a weakly supervised training method. To address the issue of too few positive samples, you might consider techniques such as data augmentation, semi-supervised learning, or leveraging pre-trained models to improve your dataset's diversity and robustness. Additionally, you could explore methods like pseudo-labeling, where the model generates labels for unlabeled data to expand the training set. If you decide to proceed with a PR, please ensure it aligns with the existing codebase and contributes effectively to the community. We look forward to your contributions!
Search before asking
Description
Hello author, I would like to add a weakly supervised training method based on YOLOV8 to address the issue of too few positive samples in my dataset. Do you have any good suggestions?
Use case
No response
Additional
No response
Are you willing to submit a PR?