The dataset curation and model analysis tool by and for the open source community
Today is the first anniversary of open-sourcing FiftyOne! We are super excited and grateful to the 40,000 early adopters that are using FiftyOne to build data-centric workflows for their computer vision projects. Thank you!
The era of model-centric machine learning is gone. FiftyOne users have grokked this shift towards the increasing importance that data plays in machine learning. Our early adopters have enabled FiftyOne to become a leading tool for dataset curation and model analysis in computer vision; its open-source nature is facilitating easy and early adoption in myriad verticals like healthcare, surveillance, advertising, and autonomous driving and across environments like individual workstations, cloud and on-premises clusters. In each of these contexts, FiftyOne is unblocking the data quality and model bottlenecks allowing users to rapidly improve performance, with an emphasis on the critical role data plays in modern AI.
The community interest, support and involvement for FiftyOne has been amazingly helpful over the last year as we have extended the data-centric capabilities to include features like Jupyter notebook support, embeddings-based analysis capabilities, tight integrations with major datasets, like Google’s Open Images and COCO, as well as tools like Lightning Flash, CVAT, Labelbox and Scale AI. This one year anniversary comes along with the FiftyOne v0.12 release, which delivers a fully optimized user interface allowing datasets to scale to hundreds of thousands of samples without a performance hit.
As we roll out FiftyOne Teams, the enterprise version of FiftyOne, we are intensely thankful to the open-source early adopters who have helped us grow the project. And, of course, the open-source version of FiftyOne is here to stay, and will remain free, forever, as we continue to enhance its functionality. The team at Voxel51 is looking forward to an exciting year ahead!
