Simplifying Classification Workflows With Ultralytics YOLOv5 v6.2

YOLOv5 is pushing the state of the art in object detection to new heights! From new classification models, training reproducibility, and Apple Metal Performance Shader (MPS) Support, to integrations with ClearML and Deci, we present to you the new YOLOv5 v6.2 release.

Link to this sectionImportant YOLOv5 Updates#

We’ve been working on improving your favorite YOLO Vision AI architecture since our latest release in February 2022. These are the most important updates in the latest YOLOv5 v6.2:

• Classification Models: YOLOv5-cls ImageNet-pretrained classification models are now available for the first time.
• ClearML logging: Integration with the open-source experiment tracker ClearML. Installing with pip install clearml will enable the integration and allow users to track every training run in ClearML. This in turn allows users to track and compare runs and even schedule runs remotely.
• GPU Export Benchmarks: Benchmark (mAP and speed) all YOLOv5 export formats with python utils/benchmarks.py --weights yolov5s.pt --device 0 for GPU benchmarks or --device CPU for CPU benchmarks.
• Training Reproducibility: Single-GPU YOLOv5 training with torch>=1.12.0 is now fully reproducible, and a new --seed argument can be used (default seed=0).
• Apple Metal Performance Shader (MPS) Support: MPS support for Apple M1/M2 devices with --device mps (full functionality is pending torch updates in pytorch/pytorch#77764).

Link to this sectionNew Classification Models#

Our main goal with this release is to introduce simple YOLOv5 classification workflows, just like our existing object detection models. The new v6.2 YOLOv5-cls models are just a start, we will continue to improve these going forward together with our existing detection models. We'd love your contributions to this effort!

This release incorporates 401 PRs from 41 contributors since our last release in February 2022. It adds Classification training, validation, prediction and export (to all 11 formats), and also provides ImageNet-pre-trained YOLOv5m-cls, ResNet (18, 34, 50, 101) and EfficientNet (b0-b3) models.

We trained YOLOv5-cls classification models on ImageNet for 90 epochs using a 4xA100 instance, and we trained ResNet and EfficientNet models alongside with the same default training settings to compare. We exported all models to ONNX FP32 for CPU speed tests and to TensorRT FP16 for GPU speed tests. We ran all speed tests on Google Colab Pro for easy reproducibility.

Link to this sectionWhat's Next From Ultralytics?#

Our next release, v6.3, is scheduled for September 2022 and will bring official instance segmentation support to YOLOv5, with a major v7.0 release later this year updating architectures across all 3 tasks - classification, detection, and segmentation.

Visit our YOLOv5 open-source GitHub repository to stay up to date and find out more about this release.