I trained a yolov8s model on my own training set，export to openvino, I cannot get correct predict results about ultralytics

[github-actions]

👋 Hello @yao-xiaofei, 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.

[glenn-jocher]

Hey there! 👋 Thanks for reaching out and providing a detailed explanation of your issue. Based on the situation you've described, it seems like setting dynamic=True during model export to OpenVINO has resolved the issue for your specific use case.

OpenVINO's dynamic shapes feature allows your model to handle variable input sizes, which might be why you're experiencing better results with dynamic=True. This setting can indeed be necessary for certain models and datasets, especially if your input images vary significantly in size or if the model has been trained on a wide range of input sizes.

Your successful use of dynamic shapes with OpenVINO suggests that this feature accommodates the variability in your data effectively. Given the comparison points you've mentioned, I'd recommend continuing to use dynamic=True for exporting your YOLOv8s model to OpenVINO format. Here’s how you can modify your export code to include that:

from ultralytics import YOLO

model = YOLO('datamatrixs.pt')
model.export(format='openvino', dynamic=True, half=False)

This configuration ensures that your model remains flexible to various input sizes, which seems to align with the successful inference results you're getting for all your images. 😊

If you have further questions or run into more issues, feel free to ask. Happy coding!