v1.0.8

add prediction parameter to annotation upload
Adds a new optional is_prediction parameter to the upload functions on projects.
passing is_prediction=True to the upload function when uploading images with annotations allows you to indicate that the annotation data you are adding is a prediction (e.g. from another model or other active learning workflow).

If the annotation is added as a prediction the image will stay in the batch or annotation job it already is in, rather than being added to the dataset for training as part of uploading the annotation data.

bug fixes
minor bug fix for properly using the OBJECT_DETECTION_URL env variable to override inference endpoint to use for object-detection models

v1.0.7

minor bug fix for properly using the OBJECT_DETECTION_URL env variable to override inference endpoint to use for object-detection models

add prediction parameter to annotation upload

Adds a new optional is_prediction parameter to the upload functions on projects.

passing is_prediction=True to the upload function when uploading images with annotations allows you to indicate that the annotation data you are adding is a prediction (e.g. from another model or other active learning workflow).

If the annotation is added as a prediction the image will stay in the batch or annotation job it already is in, rather than being added to the dataset for training as part of uploading the annotation data.

v1.0.5

This moves and renames a single method.

Version.get_pt_weights is replaced by model.download(), which makes more sense: download a model's weights is an operation that should belong to the model (version already has a download() method to download images)

V1.0.4

Adds the ability to download model weights via Version.get_pt_weights()

Fix Windows Config

In this release we support windows with new login saved configuration files.

Cheers!

Avoid Cache For Model Upload API

A small improvement to the deploy() function by avoiding caching for an API ensuring we are operating on the most up to date data.

v1.0.1

Added support for more model deploy types. Supported types are:

• YOLOv5 Object Detection
• YOLOv5 Instance Segmentation
• YOLOv7 Instance Segmentation
• YOLOv8 Object Detection
• YOLOv8 Instance Segmentation
• YOLOv8 Classification

v1.0.0

In this PR we release a few new top level methods:

.login()
.load_model()

And a cool new model method:

.webcam()

.login() will make life a lot easier, and we'll be working to propagate it throughout out docs + tutorials.

Cheers,
Team Roboflow

v0.2.34

We removed the test pypi publish from our publish action, as we often do that manually ahead of time