Turicreate

Fixes

* Fixed critical bug in logistic/linear regression with validation tracking ( 532 )

Major Updates

* Plot API with save and ability to select plot type
* Support for `list`, `dict`, `array`, `ndarray` in `.explore()`

Enhancements

* Ability to view truncated values in .explore()

Fixes

* Fix slicing issue for NDArray inside SArray (Python 3) (385)
* Make docstring consistent with current behavior (374)
* Avoiding default atof behavior for integers and floats (373)
* Fix image classifier unit tests (371)
* Consistent color for draw_bounding_boxes (369)
* Fix user facing deprecation warning for panda's to_datetime usage (368)
* Added Top Level load_sarray (516)
* user guide should use print as a function (432)
* Better error message for bad astype(...). (430)
* Remove (unused, non-functional, non-public) viz for recsys models (428)
* Better error message for unwritable temp location (427)
* Fix macOS platform tag to reflect true support (426)
* Update get/set runtime config docs and errors (423)
* Remove unnecessary Python linking (421)
* Remove references to lfs (420)
* Update supported and recommended MXNet versions (418)
* Fix single-image classifier inference on multi-gpu (417)
* Fix broken unit test for image bit depth (416)
* Add data preparation step to object detection user guide (414)
* Gitignore cleanup (408)
* updated sys.platform for python3 (404)
* Fix offset/length bug in NDArray-to-SArray (395)
* Support PNG images with non-8-bit depth (393)
* Do not use PIL for decoding images for pixel_data (390)
* Added support for file:// as valid url protocol. (387)
* Avoid ndarray to fix fatal error in activity classifier (386)

Major Updates

* Visualization support on Linux (1)
* NDArray support in SFrames
* Embedded visualizations in Jupyter/iPython notebooks (330)

Enhancements

* Single dictionary support for object detection (293 )
* Image Classifier & Image Similarity: take as input single image or SArray of images (280)
* Add validation-set statistics to resulting model for linear/logistic regression (274)

Fixes

* load_images should have a way to order the images alphabetically by path (125)
* Adds verbose flag to ImageClassifier.evaluate (281)
* csv header parsing problem with BOM UTF8 (227)
* Support palette (type P) PNG image files (278)
* Allow CUDA GPU to be used if available (227)
* Handle RGBA (4 channels) and L (1 channel) images in detector (275)

Fixes

* import turicreate fails on macOS 10.12.6 (https://github.com/apple/turicreate/issues/256)
* Miscellaneous documentation consistency fixes

Major Updates

* Support for Python 3.5 and 3.6 (2)
* Added Core ML support for sentence classifier (99)
* Added Core ML support for image similarity (242)
* Added more datasets to documentation (251)

Enhancements

* Implemented SArray.stack (https://github.com/apple/turicreate/pull/165)
* Implemented SArray.value_counts() (https://github.com/apple/turicreate/pull/172)
* Added methods to flexible_type (reinterpret_get, reinterpret_mutable_get) to make unsafe use explicit (https://github.com/apple/turicreate/pull/197)
* Published C++ Documentation at https://apple.github.io/turicreate/docs/cpp/

Fixes

* Visualization issues and improvements (https://github.com/apple/turicreate/pull/159, https://github.com/apple/turicreate/pull/166, https://github.com/apple/turicreate/pull/163)
* Core ML errors for some models trained on dict features (https://github.com/apple/turicreate/issues/98, https://github.com/apple/turicreate/issues/174)
* Topic Model perplexity computation may give different results (https://github.com/apple/turicreate/pull/243)
* String target for sentence classifier would perform feature extraction on the target (https://github.com/apple/turicreate/issues/234)
* Core ML version export type mismatch (https://github.com/apple/turicreate/pull/232)
* Sentence classifier ignored validation_set=None (https://github.com/apple/turicreate/pull/208)
* S3 Certificate Issues on Linux (https://github.com/apple/turicreate/pull/196, https://github.com/apple/turicreate/issues/57, https://github.com/apple/turicreate/issues/128)
* JSON export of array dtype (https://github.com/apple/turicreate/pull/181)
* Unable to create SFrame with non-ascii column name in Python 2 (https://github.com/apple/turicreate/issues/157)
* SArray.apply fails with UnpicklingError (https://github.com/apple/turicreate/issues/124)
* pyodbc type mismatch (https://github.com/apple/turicreate/issues/93)

Turi Create 4.0: the initial open source release!

Turi Create simplifies the development of custom machine learning models. You
don't have to be a machine learning expert to add recommendations, object
detection, image classification, image similarity or activity classification to
your app.

* **Easy-to-use:** Focus on tasks instead of algorithms
* **Visual:** Built-in, streaming visualizations to explore your data
* **Flexible:** Supports text, images, audio, video and sensor data
* **Fast and Scalable:** Work with large datasets on a single machine
* **Ready To Deploy:** Export models to Core ML for use in iOS, macOS, watchOS, and tvOS apps