v0.9.3

Summary
This release includes:

• DocTR for detecting and recognizing text
• Updates to our stream interface
• Some bug fixes and other maintenance

v0.9.2

Summary

• Bugfix parsing base64 image string when source is browser (was adding unnecessary prefix)
• Validate that equal or fewer than MAX_BATCH_SIZE images are being passed to object detection inference
• Default MAX_BATCH_SIZE to infinity
• Add batch regression tests
• Add CLI to readme
• Add generic stream object
• Add preprocess/predict/postprocess to Clip to match base interface
• Readme updates
• Landing page

v0.9.1

Summary

This release includes a new stream interface, making it easy to run latency optimized inference and run custom callbacks (documentation coming soon)

v0.9.0

Summary

This release includes:

• The new inference-cli to make starting the inference server easy and automated
• A new inference-client to as a helpful utility when interacting with the inference HTTP API
• Updates and added features to the Device Manager (enterprise feature)
• Unified model APIs so that all Roboflow models adhere to a consistent processing pipeline
• Bug fixes, maintenance

Breaking Changes:

• Some model APIs have been updated (see instance segmentation and classification)

v0.8.9

Summary

This release includes a new env var DISABLE_INFERENCE_CACHE. When set to true, internal inference caching will be disabled. Also, logging has been updated to be less verbose by default. To increase verbosity, set LOG_LEVEL=DEBUG.

v0.8.8

Summary

Contains a fix in imread/imdecode logic. Also moves logic out of version.py to fix github actions.

v0.8.7

Summary

• Abandons Pillow in favor of OpenCV for faster end to end processing
• Fixes a bug with new device management logic
• Upgrades version checking logic
• Adds env var to fix Jetson 5.1.1 images

v0.8.6

Summary
This release includes logic to detect and log if there is a newer release available. It also contains a new enterprise device manager.

v0.8.5

Summary
Contains bug fixes for configurations that use the LICENSE_SERVER setting.

v0.8.4

Summary:

• Image loading is now multi-threaded for batch requests. This should increase total FPS, especially for batch requests that include large images.
• The regression test Github action now runs on a Github actions runner.
• The extras require has been fixed for the various distribution packages