Roboflow Changelog: December 2020

Welcome to the first of our monthly changelogs where we will be cataloging our recent feature additions and improvements.

Roboflow Organize

• Added ability to rebalance train/test split
• Clarified pro plan features in-app
• Sped up thumbnail generation
• Added image upload by URL to the upload API
• Paginated the image listing pages
• New annotation format: SuperAnnotate JSON (type 2)
• New annotation format: Kaggle NFL CSV
• New annotation format: PKLot XML
• Sped up cold-start time of all cloud functions by pruning dependencies
• Fixed 31 tracked issues

Roboflow Train

• Surfaced test set metrics by class
• Validated train/test splits
• Added support for upcoming on-device inference feature
• Onboarded >100 new beta testers (reach out if interested)
• Added error handling for corrupt images
• Improved speed when loading large datasets

Roboflow Infer

• Speed improvements
• Improved usage tracking
• Created video inference script
• Improved reliability
• Outputted images now visualize boxes on the original (not preprocessed) input
• Released a sample web app
• Rigorous load testing
• On-device offering in private beta testing (reach out if interested)

Roboflow Annotate

We've been quietly beta testing our new self-serve labeling tool with a select group of users. Wide release is imminent! Stay tuned.

Other

• Launched Agriculture and Safety & Security industry pages
• New enterprise page
• Updated pricing page to be more clear
• Mentions from Augmented Startups and Machine Learning Monthly
• Cited in Nature
• 6 new public datasets
• 13 new computer vision blog posts
• 9 new computer vision videos