On-device benchmarks
Explore performance metrics of popular deep learning models.
Embedl Hub provides interactive tools and thousands of on-device benchmarks of popular deep learning models. In this guide, you will learn how to use our interactive tools to compare models’ performances and find the best model for your application.
Log in to your Embedl Hub account, and visit the Explore page. Select the task and device you’re interested in using the “Filters” panel, and compare how different models perform based on metrics such as accuracy and on-device latency.
The graph on the Explore page highlights Pareto optimal models, models with the best trade-off of your chosen metrics. It also highlights models that match your “Full-text search” queries. This is useful when you want to visualize how particular architectures perform compared to all other models.
Look for a model that fulfills your latency constraints on the target hardware and that has a high accuracy. High accuracy on a diverse dataset suggests better performance on your own dataset during fine-tuning.
When you find an ideal model, click on the model’s datapoint in the graph. This will take you to the model’s details page, where you can learn more about the model and see its performance on all tasks and devices.