Get started with 10x faster and 12x smaller YOLOv5

Neural Magic is excited to share that we’ve sparsified YOLOv5 for a 10x increase in performance and 12x smaller model files. You can now use tools and integrations linked from the YOLOv5 model page to:

  1. Reproduce our benchmarks
  2. Train YOLOv5 on new datasets to replicate our performance with your own data.

We also wrote a blog that speaks to our methodology and digs deeper into benchmarking numbers. Let us know what you think!

– Team Neural Magic

@deep-sparse-admin Any chance of browser/tfjs compatible versions of your models? This is exactly the sort of thing that the web needs, since you can make the models so small and get good performance despite the limitations of the runtime.