Exporting ONNX OBB models But the dimensions of the output results are inconsistent about ultralytics HOT 3 CLOSED

👋 Hello @hacktmz, 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.

from ultralytics.

root@zhenty-33-0:/gpu2-data/zhenty/yolov8# yolo export model=runs/obb/train/weights/best.pt format=onnx opset=15 imgsz=640
Ultralytics YOLOv8.2.10 🚀 Python-3.8.10 torch-1.14.0a0+44dac51 CPU (Intel Xeon Platinum 8255C 2.50GHz)
YOLOv8s-obb summary (fused): 187 layers, 11411958 parameters, 0 gradients, 29.4 GFLOPs

PyTorch: starting from 'runs/obb/train/weights/best.pt' with input shape (1, 3, 640, 640) BCHW and output shape(s) (1, 6, 8400) (44.9 MB)

ONNX: starting export with onnx 1.12.0 opset 15...
========== Diagnostic Run torch.onnx.export version 1.14.0a0+44dac51 ===========
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ========================

ONNX: export success ✅ 2.7s, saved as 'runs/obb/train/weights/best.onnx' (43.7 MB)

Export complete (6.4s)
Results saved to /gpu2-data/zhenty/yolov8/runs/obb/train/weights
Predict: yolo predict task=obb model=runs/obb/train/weights/best.onnx imgsz=640
Validate: yolo val task=obb model=runs/obb/train/weights/best.onnx imgsz=640 data=/gpu2-data/dataset/ocrdata20240430/obb.yaml
Visualize: https://netron.app
💡 Learn more at https://docs.ultralytics.com/modes/export
root@zhenty-33-0:/gpu2-data/zhenty/yolov8# yolo export model=runs/obb/train/weights/best.pt format=onnx opset=15 imgsz=1024
Ultralytics YOLOv8.2.10 🚀 Python-3.8.10 torch-1.14.0a0+44dac51 CPU (Intel Xeon Platinum 8255C 2.50GHz)
YOLOv8s-obb summary (fused): 187 layers, 11411958 parameters, 0 gradients, 29.4 GFLOPs

PyTorch: starting from 'runs/obb/train/weights/best.pt' with input shape (1, 3, 1024, 1024) BCHW and output shape(s) (1, 6, 21504) (44.9 MB)

ONNX: starting export with onnx 1.12.0 opset 15...
========== Diagnostic Run torch.onnx.export version 1.14.0a0+44dac51 ===========
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ========================

ONNX: export success ✅ 2.9s, saved as 'runs/obb/train/weights/best.onnx' (43.8 MB)

Export complete (8.2s)
Results saved to /gpu2-data/zhenty/yolov8/runs/obb/train/weights
Predict: yolo predict task=obb model=runs/obb/train/weights/best.onnx imgsz=1024
Validate: yolo val task=obb model=runs/obb/train/weights/best.onnx imgsz=1024 data=/gpu2-data/dataset/ocrdata20240430/obb.yaml
Visualize: https://netron.app
💡 Learn more at https://docs.ultralytics.com/modes/export
root@zhenty-33-0:/gpu2-data/zhenty/yolov8#

from ultralytics.

It looks like the output dimensions differ based on the imgsz used during the model export process. When the model is trained with imgsz=640 and exported with the same, the dimensions are adjusted accordingly which differs from when exported with imgsz=1024.

It's important to maintain consistency between the training and exporting image sizes to ensure the output dimensions align with expectations. If you need the output dimensions to match a specific size, consider retraining the model with that exact imgsz or adjust the imgsz during export to match your trained model settings.

If you have further issues or questions, feel free to reach out! 🚀

from ultralytics.