Green check
Link copied to clipboard

YOLOv5 Just Got Stronger in v6.1!

We strive to make machine learning deployment simpler, better, faster!

Facebook logoTwitter logoLinkedIn logoCopy-link symbol

You shouldn't wait much longer! From TensorRT, Edge TPU and OpenVINO support, to retrained models at --batch-size 128 and a new default one-cycle linear LR scheduler, we present to you YOLOv5 v6.1 - much simpler, faster, better, stronger!

YOLOv5 v6.1 release

Important Updates

Ever since our last release in October 2021, we've been working on improving your favorite YOLO Vision AI architecture. From bug fixes to new features, these are the most important enhancements in the latest YOLOv5 v6.1 release:

  • TensorRT support: TensorFlow, Keras, TFLite, TF.js model export now fully integrated using python --include saved_model pb tflite tfjs (#5699 by @imyhxy).
  • Tensorflow Edge TPU support ⭐ NEW: New smaller YOLOv5n (1.9M params) model below YOLOv5s (7.5M params), exports to 2.1 MB INT8 size, ideal for ultralight mobile solutions (#3630 by @zldrobit).
  • OpenVINO support: YOLOv5 ONNX models are now compatible with both OpenCV DNN and ONNX Runtime (#6057 by @glenn-jocher).
  • Export Benchmarks: Benchmark (mAP and speed) all YOLOv5 export formats with python utils/ --weights Currently operates on CPU, future updates will implement GPU support (#6613 by @glenn-jocher).
  • Hyperparameters: minor change.
  • hyp-scratch-large.yaml lrf reduced from 0.2 to 0.1 (#6525 by @glenn-jocher).
  • Training: Default Learning Rate (LR) scheduler updated.
  • One-cycle with cosine replace with one-cycle linear for improved results (#6729 by @glenn-jocher).

YOLOv5 now officially supports 11 different formats, not just for export but for inference (both and PyTorch Hub), and validation to profile mAP and speed results after export. Check the list of supported models in the latest YOLOv5 for yourself:

TensorFlow SavedModel
TensorFlow GraphDef
TensorFlow Lite
TensorFlow Edge TPU

Together for Everyone's AI

Inspired by the Olympic spirit, we at Ultralytics believe that the most important thing is not to win but to take part. We acknowledge our YOLOv5 family has taken an enormous part in both our triumphs and struggles. This release has been brought to you by 271 PRs from 48 new contributors. Our shared vision to make AI accessible for everyone has brought us together to create the most accurate ML models.

Want to make part of the best AI in the world?

We are always looking for talents to join our team or contribute to our open-source projects!

From AI Enthusiasts to the Most Popular Object Detection of 2022

This month our Ultralytics/YOLOv5 repository has surpassed Joseph Redmon's pjreddie/darknet YOLOv3 by the total number of GitHub stars, now counting over 22.4k. We are humbled by the opportunities ahead in the Vision AI space for YOLOv5 and beyond. It is our pleasure to carry the You Only Look Once legacy forward.

Visit our YOLOv5 open-source GitHub repository to find out more detail regarding this release and get into YOLO object detection.

Or even better...

Find out how you can do all YOLO-related magic with no code! Ultralytics HUB will get you into Computer Vision from scratch with a few clicks of a button!

Let’s build the future
of AI together!

Begin your journey with the future of machine learning

Read more in this category