I am thrilled to share the release of Pipeless 1.0. A new version of Pipeless that was rewritten from scratch to bring significant improvements in performance, flexibility, and development experience.
Pipeless is an open-source computer vision framework that ships everything you need to build and deploy computer vision applications really fast.
As you will see below, the amount of changes in this new version is significant. We changed from how you write code to how you configure streams. However, those changes led to an incredible user experience.
We have redesigned Pipeless taking into account all your feedback to create a new internal architecture that is much more powerful and flexible, and that can support most of your use cases.
Pipeless 1.0 is a significant step forward in our initial vision of bringing the development experience from serverless frameworks to embedded computer vision.
You can get started with Pipeless 1.0 in just seconds following the getting started guide.
- ⚡ 8x faster using just 20% of the resources. Moved to a single lightweight component approach from the former multi-component approach.
- ✨ Multi-stream support. You can now process any number of streams within a single Pipeless instance.
- 👩💻 Migrated from Python to Rust with multi-language support. You can now write processing hooks in many programming languages, including Python, Rust, and more to come.
- 🛠️ Dynamic stream configuration. No more config files. Add, edit, and remove streams instantly via the CLI or a REST API without restarting Pipeless.
- 🕸️ Highly parallelized. Your code is executed in a highly parallelized environment, reaching lightning speeds and saving you headaches. You write simple functions, Pipeless takes care of parallelizing its execution.
- ⛓️ Chain any number of processing steps. Those can include inference with different CV models, simple RGB modifications, or whatever you need.
- 📁 Structured and highly reusable user code. Pipeless now helps you create well-structured code based on filesystem routing that is highly reusable and parallelizable.
Get started with Pipeless 1.0 in just seconds by following this guide.