v0.5.2

What's Changed

This release:

• Add the possibility for the asynchronous requests to not send the request to fetch results.
• Improve the sam2 video support by compiling the predictor and accepting video files either as base64, a url, a directory or a list of frames that can be base64 or urls.
• Add examples to the documentation.

Commits

• Improve asynchronous calls by authorizing empty requests by @juliendenize in #30
• fix getting-started link by @BertrandRenault in #31
• Improve sam2 video by accepting file/base64, folder or list of files/base64 by @juliendenize in #32
• Add examples to the documentation by @juliendenize in #33

Full Changelog: v0.5.1...v0.5.2

v0.5.1

What's Changed

Add the Image zero shot detection task.

Also add possibility to make the PixanoInferenceClient work with async calls.

• Add transformers image zero shot detection by @juliendenize in #28
• Feat/client async by @juliendenize in #29

Full Changelog: v0.5.0...v0.5.1

v0.5.0

What's Changed

Pixano-Inference takes now advantage of Redis and Celery to handle requests asynchronously.

The client for now has to loop to retrieve results.

• Add Redis, celery and rework the API and client by @juliendenize in #27

Full Changelog: v0.4.1...v0.5.0

v0.4.1

What's Changed

Corrected typos in the doc and some minor fixes.

New Contributors

• @BertrandRenault made their first contribution in #26

Full Changelog: v0.4.0...v0.4.1

v0.4.0

Pixano-Inference 0.4.0

This is the official release of the 0.4.0 Pixano-Inference.

There is a complete refactor of the library that has now these features:

• A REST API to deploy and call inference models from providers
• Support to Transformers, Sam2 and vLLM providers
• Support to the image mask generation and video mask generation task
• Support to the image text conditional generation task

We will now focus on improving the library to support more tasks and providers, deploy the models efficiently and handle the requests via a queue (for now requests are discarded if the model is busy).

Please note that this is a Work In Progress and a lot of functionalities are experimental and subject to API change.

What's Changed

• Refactor Pixano-inference by @juliendenize in #20
• Add minimal docs by @juliendenize in #21
• Remove sam2 optional dependency for pypi and fix asserts by @juliendenize in #22
• Add py3.13 by @juliendenize in #23
• V0.4.0b2 by @juliendenize in #24
• Feat/improve vllm by @juliendenize in #25

New Contributors

• @juliendenize made their first contribution in #20

Full Changelog: v0.3.1...v0.4.0

v0.4.0b2

What's Changed

Add vLLM provider, a client and perform some improvements for text image conditional generation as well as image mask generation.

• V0.4.0b2 by @juliendenize in #24

Full Changelog: v0.4.0b1...v0.4.0b2

v0.4.0b1

What's Changed

• Add py3.13 by @juliendenize in #23

Full Changelog: v0.4.0b0...v0.4.0b1

v0.4.0b0

New version incoming !

We are in the process of transforming Pixano-Inference to handle more efficiently models and data for Pixano. We will support transformers, sam2, and later well know providers such as OpenAI, Mistral, ...

Stay tuned !

What's Changed

• Refactor Pixano-inference by @juliendenize in #20
• Add minimal docs by @juliendenize in #21
• Remove sam2 optional dependency for pypi and fix asserts by @juliendenize in #22

New Contributors

• @juliendenize made their first contribution in #20

Full Changelog: v0.3.1...v0.4.0b0

Pixano Inference 0.3.1

Under active development, subject to API change

Install this version of Pixano Inference with pip install pixano-inference~=0.3.1

Added

• Add new GroundingDINO model for semantic segmentation with text prompts (#6)

Changed

• Update README badges with PyPI release

Fixed

• Remove top-level imports for GitHub models to prevent import errors (#6)
• Fix preannotation with SAM and MobileSAM (#6)
• Add type hints for Image PixanoType (#6)
• Update Pixano requirement from 0.5.0 beta to 0.5.0 stable

Pixano Inference 0.3.0

Under active development, subject to API change

Install this version of Pixano Inference with pip install pixano-inference~=0.3.0

Added

• Publish Pixano Inference on PyPI
• Add new MobileSAM model as a lighter alternative to SAM (#2)
• Add GitHub actions to format and lint code
• Add GitHub action to publish docs and PyPI package (#9)
• Add issue and pull request templates on GitHub repository
• Add CONTRIBUTING.md for installation information and contribution guidelines

Changed

• Breaking: Remove SAM and MobileSAM dependencies to allow publishing to PyPI (#14)
• Breaking: Update to Pixano 0.5.0
• Breaking: Update InferenceModel id attribute to model_id to stop redefining built-in id
• Breaking: Update submodule names to pytorch, tensorflow, and github
• Update README with a small header description listing main features and more detailed installation instructions
• Generate API reference on documentation website automatically
• Add cross-references to Pixano, TensorFlow, and Hugging Face Transformers in the API reference
• Update documentation deployment (#9)

Fixed

• Fix links to Pixano notebooks
• Fix internal cross-references in the API reference of the documentation website
• Update deprecated GitHub actions (#11)
• Prevent CUDA installation in lint GitHub action action
• Fix ignored members in linting CI configuration
• Specify black version in formatting CI configuration
• Fix GitHub version and documentation links in README