Enhancement: Add ONNX export and benchmarking for inference optimization

[Ayushd172005]

Is there an existing issue for this?

• I have searched the existing issues

What happened?

PictoPy currently lacks a standardized approach to evaluate and optimize inference performance.

This issue proposes:

• Exporting models to ONNX format
• Benchmarking inference performance (PyTorch vs ONNX Runtime)
• Establishing a foundation for hardware-accelerated inference

This is being addressed in PR #1248.

Record

• I agree to follow this project's Code of Conduct

[github-actions]

👋 Thanks for opening this issue! The maintainers will review it shortly. Till then, do not open any PR. This is a strict rule we like to follow around here :)

[JITENDRAGAHERWAR03]

Hello Maintainers 👋
I would like to work on this enhancement: ONNX export and benchmarking for inference optimization.
I have analyzed the current limitation in PictoPy regarding the absence of a standardized inference benchmarking pipeline. To address this, I propose implementing:
Exporting trained models (e.g., YOLOv8) to ONNX format
Running inference using ONNX Runtime
Benchmarking performance (latency, inference time, FPS) against PyTorch
Creating a structured comparison for optimization
This enhancement will help improve inference speed, reduce latency, and enable better hardware utilization.
I have experience working with:
Python, PyTorch, and Machine Learning
Data Analysis & Performance Optimization
Building structured and scalable solutions
I am confident in implementing this feature along with proper documentation and benchmarking results.
Looking forward to your feedback and approval to proceed 🚀
Thank you!