Voxel51
End-to-end multimodal data platform for physical AI
Start new thread

FiftyOne - Accelerate your machine learning experimentation

by
•
Good ML models require rapid iteration. FiftyOne helps you visualize and explore your datasets, identify key scenarios, and evaluate your models’ performance. Spend less time wrangling and more time on the science of building better ML models.

Add a comment

Replies

Best
Maker
Hey there! I’m Brian, Co-Founder and CTO of Voxel51 (voxel51.com). I’m excited to share FiftyOne, our open source tool that accelerates machine learning experimentation. Nothing hinders the success of ML systems more than poor quality data. And without the right tools, improving a ML model can be a time-consuming and inefficient process. We built FiftyOne to help ML engineers and scientists spend dramatically less time on ad hoc experimentation and instead uncover the insights needed to build better datasets and better models. With FiftyOne, you can: - Curate diverse, high quality datasets - Visualize and evaluate annotations and model outputs - Search, filter, and explore scenarios - Identify failure modes in your models We believe developer tools are most impactful when they are open and transparent. That’s why we released FiftyOne open source on GitHub! We want everyone to have the best tools for rapidly iterating ML models. We’re excited to hear your thoughts and see the amazing ML products that are built using FiftyOne! Drop us a comment here or email me directly at .
Does it make use of the M1's ML cores?
Maker
FiftyOne is a dataset analysis library that sits on top of ML frameworks like TensorFlow and PyTorch. So you'll see performance boosts on M1 cores whenever the underlying ML framework allows it. For example, with TF 2.4, users are seeing 5-10x performance gains: . If you'd like to try it out yourself, you can load a model from FiftyOne's Model Zoo and add predictions to a dataset with just a couple lines of code. Check out the link below to learn more: Cheers, Brian