Creating a CVOps Platform: Picsellia’s Latest Release is Out

Creating a CVOps Platform: Picsellia’s Latest Release is Out

Our Road to CVOps

Our team has been hard at work making some big changes to make Picsellia the most complete European end-to-end CVOps platform—in short, MLOps applied to computer vision.

Our new release incorporates the latest tools and features to manage your entire MLOps pipeline in one single place, especially optimized for computer vision.

Until now, Picsellia only centralized dataset management, experiment tracking and model training tools in a single platform to assist data scientists in their research phase. However, since the Deep Learning market has evolved beyond just research, we wanted to go a step further and create a stack that covered the entire MLOps pipeline for computer vision projects.

What is New in Picsellia?

It’s time to put Computer Vision models in production!

Through an easy-to-use and powerful ecosystem, our end-to-end CVOps solution simplifies and centralizes workflows, for research to production, while promoting collaboration across distributed teams.

Today, Picsellia does not only cover the research stage, but also the production pipeline—and it’s up to you whether to make full use of the platform as a whole, or integrate it with other existing tools.

Here are some of the new features:

1. Serverless Model Deployment

Put your computer vision model in production in a minute. Now you can jump from a trained experiment to a model put in production in a robust and scalable infrastructure. Our serverless deployment solution comes with an out-of-the box monitoring suite.

No alt text provided for this image

We handle all the heavy lifting for you. In just a click or few lines of code, you get an API endpoint to call your model at scale.

2. Model Monitoring

No alt text provided for this image

Deep Learning without monitoring is like driving a car without checking the fuel gauge. Picsellia users can now benefit from an entire drift-detection toolbox in seconds. Monitor your computer vision models performance in real time. Easily identify and fix potential failure cases —i.e.: data drift cases—and make sure that everything is running smoothly, at all times.

3. Automated Workflows

Building robust computer vision models is a continuous process that takes time and patience. Small-sized but quality data is always a good starting point to build a performing model, but it doesn’t have to stay that way forever. Big-sized and quality data is possible to achieve once you create an intelligent feedback loop and monitoring.

No alt text provided for this image

Picsellia's automated pipelines allow you to trigger trainings and alerts every time relevant data enters the feedback loop. This way, you can increase the size of your dataset with interesting data, like edge cases or under-represented classes.

You can then set up some rules to automatically deploy your new models based on their validation performances. This way, you can put your whole MLOps pipeline on auto-pilot.

What’s Next for Picsellia?

Our vision at Picsellia has always been to support innovative Deep Learning teams across the globe, and to increase the adoption of Computer Vision in all industries.

Given the rapid evolution of the market towards real-world applications, new challenges have emerged, some of them being data-drift and the need of continuous monitoring. Today, our mission is to create pipelines integrating human-in-the-loop, and to allow businesses to have a smart understanding of their models to drive value to people.

If you’d like to learn more about Picsellia, schedule a quick demo with our Sales team!

Start managing your AI data the right way.

Request a demo

Recommended for you:

french language
english language