5d in Onnx about yolov5

[github-actions]

👋 Hello @Manueljohnson063, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

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.

Requirements

Python>=3.8.0 with all requirements.txt installed including PyTorch>=1.8. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 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 YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training, validation, inference, export and benchmarks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

Introducing YOLOv8 🚀

We're excited to announce the launch of our latest state-of-the-art (SOTA) object detection model for 2023 - YOLOv8 🚀!

Designed to be fast, accurate, and easy to use, YOLOv8 is an ideal choice for a wide range of object detection, image segmentation and image classification tasks. With YOLOv8, you'll be able to quickly and accurately detect objects in real-time, streamline your workflows, and achieve new levels of accuracy in your projects.

Check out our YOLOv8 Docs for details and get started with:

pip install ultralytics

[glenn-jocher]

Hi there,

Thank you for reaching out! It sounds like you're encountering an issue with a 5D tensor in your ONNX model that your hardware does not support. To address this, you can reshape the tensor to a lower dimension, such as two 4D tensors, directly within the ONNX model.

Here's a general approach to achieve this using ONNX's built-in operations:

1. Split the 5D tensor: Use the Split operation to divide the 5D tensor into two 4D tensors.
2. Reshape the tensors: Apply the Reshape operation to each of the resulting tensors to ensure they fit the required dimensions.

Here's an example of how you might do this in Python using the onnx library:

import onnx
from onnx import helper, numpy_helper

# Load your ONNX model
model = onnx.load('your_model.onnx')

# Define the split operation
split_node = helper.make_node(
    'Split',
    inputs=['input_tensor'],
    outputs=['output_tensor_1', 'output_tensor_2'],
    axis=0,  # Adjust the axis based on your tensor's shape
    split=[2, 3]  # Adjust the split sizes based on your tensor's shape
)

# Define the reshape operations
reshape_node_1 = helper.make_node(
    'Reshape',
    inputs=['output_tensor_1', 'shape_tensor_1'],
    outputs=['reshaped_tensor_1']
)

reshape_node_2 = helper.make_node(
    'Reshape',
    inputs=['output_tensor_2', 'shape_tensor_2'],
    outputs=['reshaped_tensor_2']
)

# Add the nodes to the graph
model.graph.node.extend([split_node, reshape_node_1, reshape_node_2])

# Save the modified model
onnx.save(model, 'modified_model.onnx')

Make sure to adjust the axis and split parameters based on the specific shape of your 5D tensor. Additionally, you will need to define the shape_tensor_1 and shape_tensor_2 appropriately to reshape the split tensors into 4D tensors.

If you encounter any issues or need further assistance, please ensure you are using the latest version of the ONNX package and YOLOv5. Feel free to share any error messages or additional details, and we'll be happy to help further!

[Manueljohnson063]

Many Many thanks for the quick replay ,
In my case the input to the reshape node is 4d and output is 5d ,

,
Many many thanks in advance .

@glenn-jocher

[glenn-jocher]

@Manueljohnson063 hi there,

Thank you for the additional details and the kind words! 😊

Given that your input to the reshape node is a 4D tensor and the output is a 5D tensor, we need to adjust our approach to ensure compatibility with your hardware. Here’s a refined solution to reshape the 5D output into a more manageable form, such as two 4D tensors.

Solution Overview

1. Reshape the 5D tensor back to a 4D tensor: This can be done by merging one of the dimensions.
2. Split the resulting 4D tensor: If needed, split the 4D tensor into two separate 4D tensors.

Example Code

Here’s an example using the onnx library in Python:

import onnx
from onnx import helper, numpy_helper

# Load your ONNX model
model = onnx.load('your_model.onnx')

# Define the reshape operation to convert 5D to 4D
reshape_node = helper.make_node(
    'Reshape',
    inputs=['input_tensor'],
    outputs=['reshaped_tensor'],
    shape=[-1, 4, 4, 4]  # Adjust the shape based on your tensor's dimensions
)

# Optionally, define a split operation if you need to split the 4D tensor
split_node = helper.make_node(
    'Split',
    inputs=['reshaped_tensor'],
    outputs=['output_tensor_1', 'output_tensor_2'],
    axis=0,  # Adjust the axis based on your tensor's shape
    split=[2, 2]  # Adjust the split sizes based on your tensor's shape
)

# Add the nodes to the graph
model.graph.node.extend([reshape_node, split_node])

# Save the modified model
onnx.save(model, 'modified_model.onnx')

Notes

• Adjust the shape parameter in the Reshape node to fit your specific tensor dimensions.
• If you need to split the 4D tensor further, adjust the axis and split parameters in the Split node accordingly.

Verification

Please ensure you are using the latest versions of the ONNX package and YOLOv5 to avoid any compatibility issues. If you encounter any errors or need further assistance, feel free to share more details, and we’ll be happy to help!

Thank you for your patience and collaboration. The YOLO community and the Ultralytics team are always here to support you!