TFLite NMS is agnostic only?

[titanium-cranium]

@glenn-jocher @Y-T-G Excited to see the NMS feature added. I was wondering, however, whether the TFLite nms is agnostic only?

I might be reading this incorrectly but I think that's what is going on here:

ultralytics/ultralytics/engine/exporter.py

Line 417 in 9d8c2fe

f[7], _ = self.export_tflite(keras_model=keras_model, nms=False, agnostic_nms=self.args.agnostic_nms)

[glenn-jocher]

👋 Hello @titanium-cranium, thank you for your interest in Ultralytics 🚀! We’re glad to see your enthusiasm for the new NMS feature 🎉.

We recommend checking out the Docs for detailed information on YOLO features and implementations. You can also explore the Export Tutorials for guidance on exports, such as TFLite.

If this is a 🐛 Bug Report, please provide a minimum reproducible example showcasing the behavior you’ve observed to help us better understand and debug the issue.

For further clarification regarding your question or if you’re experimenting with exports and TFLite, feel free to share more specifics about your workflow. This could include code snippets, configurations, or any logs that might offer additional context. Providing these details will allow us and the community to offer more precise assistance 👨‍💻.

Meanwhile, you can join the Ultralytics community to continue this discussion and explore more with like-minded folks:

• Real-time chat on Discord 🎧
• In-depth discussions on Discourse
• Share experiences on our Subreddit

Upgrade

To ensure you’re working with the latest features and fixes, upgrade to the latest ultralytics package, including all requirements, using:

pip install -U ultralytics

Ensure your environment uses Python>=3.8 and PyTorch>=1.8.

Environments

For seamless experimentation, YOLO can be run in any of the following verified environments, each pre-configured for efficient deployment:

• Notebooks with free GPU:
  • 
  • 
  • 
• Google Cloud Deep Learning VM – GCP Quickstart Guide
• Amazon Deep Learning AMI – AWS Quickstart Guide
• Docker Image – Docker Quickstart Guide

Status

The Ultralytics CI system runs scheduled and per-commit checks to ensure the integrity of all YOLO Modes and Tasks across macOS, Windows, and Ubuntu.

This is an automated response 🤖, but an Ultralytics engineer will follow up with you soon 🔍. Stay tuned!

[Y-T-G]

It isn't. It supports both agnostic and class-wise NMS.

[titanium-cranium]

Is there a particular reason no CLI argument is being passed through to allow this?

[Y-T-G]

self.args.agnostic_nms can be both True or False

[titanium-cranium]

Sorry, I may be being completely dense here but the nms arg is harcoded to False no?

[Y-T-G]

This seems like old code, not related to the NMS that was added recently. The args here doesn't seem like they're used.

[titanium-cranium]

Thanks @Y-T-G I'll trace it back to see how to use it properly.