Output shape of [1,5,2100] about ultralytics HOT 4 CLOSED

👋 Hello @Bello-dev, 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.

Hello,

The output shape [1, 5, 2100] you're observing after exporting your model to ONNX seems to be related to the number of anchors and the grid size used during the training of your model. The standard output shape [1, 5, 8400] typically corresponds to a different configuration, possibly with more or larger grid sizes.

If you've previously trained models with only one class and are now working with a different setup or a different number of classes, this might affect the output shape due to changes in the configuration settings such as anchor boxes, grid sizes, or the number of classes.

Please ensure that your model configuration during training matches your expectations and intended use case. You might want to review the configuration file used for training (*.yaml) to verify that all parameters are set correctly according to your current needs.

If everything seems correct and you're still facing issues, please provide more details about your training configuration and we can look deeper into it.

from ultralytics.

i didn't change any config i just run the command as always
yolo detect train data=data.yaml model=yolov8n.pt epochs=100 imgsz=320 device=0 batch=64

using the data.yaml

train: ../train/images
val: ../valid/images
test: ../test/images

nc: 1
names: ['boat']

from ultralytics.

@Bello-dev hello,

Thank you for providing the details of your training command and configuration. From what you've shared, it seems like your setup is consistent with training a model for a single class ('boat').

The output shape [1, 5, 2100] you're seeing is likely due to the specific grid sizes and anchor configurations used by the YOLOv8n model for the image size of 320x320. This configuration can result in different numbers of output predictions, which seems to be the case here.

If you haven't modified any other settings like the anchor boxes or the model architecture itself, the output shape you're observing is expected based on the internal calculations of the model for the given input size and the number of classes.

If you have further questions or need more assistance, feel free to ask. Happy to help! 😊

from ultralytics.