PerceptiLabs Fact Sheet
What is PerceptiLabs?
PerceptiLabs is a machine learning company that has developed a
next-generation ML tool for a new visual way to build machine
learning models.
PerceptiLabs’ tool includes:
● PerceptiLabs Visual Modeling Tool
● A code editor and Jupyter Notebooks view
● PerceptiLabs Model Hub
What is the PerceptiLabs tool? Why is it unique?
PerceptiLabs is a dataflow driven, visual API for TensorFlow, carefully
designed to make machine learning (or deep learning) modeling as
intuitive as possible. The visual modeler has the best of all worlds -the
flexibility of visual modelling, the ability to develop in pure code, and
some of the automation of AutoML. This gives you a faster way to
build out models, with more transparency into how your model is
architected and a view into how it performs.
How does PerceptiLabs make model building faster and
easier?
The drag and drop UI makes it easy for beginners to get started and
provides a rich set of components giving you all of the functionality
needed to build and train virtually any ML model.
How does PerceptiLabs make model building flexible?
The drag and drop user interface is dataflow driven and t he advanced
features make model building flexible yet comprehensive for
researchers and experts alike. You can choose the way you want to
work. The UI lets you toggle between the visual modeler and the code
editor. You can also view and export your model as a Jupyter
Notebook.
How does PerceptiLabs advance explainability in model
development?
The unique UI let’s you see what’s going on inside your model as you
build, train and test it. See real-time analytics in every operation and
variable, and granular previews of output from each model
component.
What automation is provided?
The degree of automation provided includes: I/O shape fitting, auto
generated hyperparameter settings, model templates and auto
generated visualizations. Automated distribution of training/workloads
over all available GPUs is also available in Team and Enterprise
versions.
We already use TensorFlow, why would I want to use
PerceptiLabs?
PerceptiLabs acts as a layer on top of TensorFlow, making that
development more efficient and streamlined and more transparent.
What are some of the common types of models I can build?
You can build any type of model you like. PerceptiLabs also includes a
number of templates that make it more efficient to build simple models
like Linear Regression and Image Classification.
I need to build complex models. Can I do that with
PerceptiLabs?
Yes. The power of PerceptiLabs is in the unique UI that makes it easy
and intuitive to build any state-of-the-art model architecture. We do
have templates to more easily build and train complex models like
GANs, reinforcement learning, and object detection.
How can I get started? What is the cost?
Developers can start building a model with our free version that’s
installed as a Python package. Just run: pip install
perceptilabs and then perceptilabs and wait for
PerceptiLabs to open in your browser.
We work in a team with many models—will this work for us?
Yes—we have a Team version where you can install PerceptiLabs
Docker version. In both the Team and Enterprise versions, the Model
Hub helps you to manage concurrent development of models across
team members.
Do you have an Enterprise version?
Yes. We have partnered with Red Hat OpenShift for enterprise
deployment.
Where to learn more:
Visit our website at www.perceptilabs.com
Or contact martin.i@perceptilabs.com
