Development Tools

Leverage the power of Ergo to provide richer user experiences.

Ergo processors make it possible to integrate large, advanced neural networks into edge devices, run multiple networks at once, and use multimodal inputs, all while drawing only milliwatts of power. To help you take advantage of these capabilities, Perceive provides a comprehensive set of development tools.

Percipio Development Platform

Built on PyTorch and PyTorch Lightning, the Percipio development platform produces highly efficient real-time neural network models for an Ergo processor. Percipio uses proprietary, loss-aware compression algorithms to compress, train, and compile models for deployment on Ergo.

  • Can be run in the cloud, on-premises, or on a workstation (minimum system requirements apply)
  • Provides PyTorch-like APIs for ease of use during development and training
  • Provides metrics and logging facilities to aid in model development and compression
  • Uses a command line interface for simple access regardless of installation location

Perceive Model Zoo


Start with our model zoo or bring your own – Percipio accepts a variety of models, including CNNs, RNNs, LSTMs, GRUs, and Transformers. Designed to provide developers with a head start, the Perceive Model Zoo is a collection of off-the-shelf models with Percipio recipes that are already ported and optimized to run on Ergo chips.

Using our models is not required – Percipio also enables you to bring your own model and optimize it for Ergo. We’d be happy to discuss your needs and help you assess the compatibility of your model.

Perceive Software Development Kit (SDK)

The Perceive SDK provides the tools you need to use your neural networks on an Ergo processor. With the SDK, you can build applications for both the chip and the host system to receive data from your sensors, run inferences, and use those inferences to provide device features. It includes:

  • Precompiled libraries for Ergo
  • APIs for software integration
  • APIs for system application development
  • Sample application source code.
The Ergo SDK support firmware development (embedded applications running on the chip) to run inferences on sensor data and pass them to the host. The System SDK enables developers to build applications that leverage these inferences to provide features for the host device.

Ergo Evaluation Board

Perceive offers evaluation boards for both Ergo and Ergo 2. Ergo Evaluation Boards are small boards that contain an Ergo or Ergo 2 device and interfaces for various sensors such as MIPI cameras and audio microphones. The boards’ form factor is designed to plug into a Raspberry Pi-type computer. An Ergo evaluation board can help test and refine applications as you build your final product.

Download the Development Tools Overview