Bounding box cannot fit

[irfanamal]

Search before asking

• [X] I have searched the YOLOv8 issues and discussions and found no similar questions.

Question

I am trying to create field detection of passport using pretrained yolov8n. I have dataset consisting of 1000 samples of well-labeled data. After fine tuned the pretrained model on my dataset, I found that it has good mAP50-95 on every label except MRZ line one. Do you know why and how to fix it?

I also attach the validation result. Top one is the ground truth image, the middle one is the prediction, and the bottom one is the validation score. Note: the dataset is synthetic, so it is fake except for the face. Therefore I have to censored it.

Additional

No response

[github-actions[bot]]

👋 Hello @irfanamal, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[Y-T-G]

What imgsz are you using?

[irfanamal]

I'm using imgsz = 640. In general, I used default hyperparameters, except for lr0, lrf, and batch which are 1e-3, 1e-6, and 32 respectively

[Y-T-G]

lrf shouldn't be that low. lrf stands for learning rate factor, not final learning rate.

final_lr = lr0 * lrf

[irfanamal]

I see... I'll try to use the default (lr0: 0.01, lrf: 0.01) and give the update

[pderrenger]

Great! Using the default learning rate settings should help. Please let us know how it goes after you make the adjustments. If the issue persists, we can explore other potential solutions.

[irfanamal]

The issue still persists

[pderrenger]

Thank you for the update. Since the issue persists, please ensure you are using the latest version of the Ultralytics package. If the problem continues, consider checking the class distribution in your dataset and ensuring that the MRZ line one class is well-represented. Additionally, you might want to experiment with different augmentation techniques to improve the model's ability to generalize. If you need further assistance, feel free to ask.

[irfanamal]

I tried bigger imgsz and it works.

[pderrenger]

Glad to hear that increasing the image size resolved the issue! If you encounter any further problems or have additional questions, feel free to reach out.