What are the differences between Piccolo AI and SensiML Analytics Studio

Technical Discussion
1

Just wondering how the open-source and equivalent paid subscription applications differ from each other. Can anyone explain?

2

SensiML Analytics Studio, as noted, is SensiML’s paid subscription SaaS (software-as-a-service) means of offering AutoML-based AI/ML inference model development to developers seeking to build for microcontroller and IoT edge device targets. Analytics Studio comes in various service tiers and is priced by tier and optional add-ons. Those interested in learning more about SensiML’s paid SaaS plans can find details here: https://sensiml.com/plans/.

SensiML Analytics Studio is a cloud-hosted service whereby users can setup their own accounts to access and use the latest version of SensiML’s AutoML engine for IoT edge devices. It provides convenience, fully managed application access, cloud project data storage and synchronization for multi-user teams, and direct support from SensiML’s development team.

SensiML Analytics Studio also supports the broadest collection of target devices with an extensive list of processor architectures as well as the option to quickly generate working binary firmware for quick testing of ML models on a variety of evaluation and prototyping boards using console, BLE, and WiFi TCP/IP model output.

SensiML Piccolo AI is a fully open-source version of SensiML’s AutoML intended to inspire users to use, extend, enhance, and contribute additional capabilities for AutoML targeting IoT edge devices and microcontrollers. Piccolo AI is completely free to use and has substantially similar capabilities to the paid Analytics Studio SaaS application with the following notable differences:

  1. Piccolo AI is intended for local execution on your own PC or server system and is not offered as a cloud service. You are fully responsible for installing and managing the application. With that, of course, comes the flexibility to extend the functionality of the application and customize the code to suit your own purposes, subject to the AGPLv3 license terms and conditions.

  2. Users must download the source code and compile the application themselves using the getting started guide accessible at sensiml.org or github.com/sensiml/piccolo. As a server application intended to run within a Docker container, a higher level of familiarity with typical software development tools is required to successfully get Piccolo AI up and running than Analytics Studio.

  3. Piccolo AI does not offer EVB-specific inference model binaries for rapid testing of models by direct flashing of bin/hex files directly to the target board. You must instead generate C source code or object code and then create your own wrapper application, compile within the target embedded development environment, and then flash your board with the resulting firmware. As a developer-focused OSS project, Piccolo AI is geared toward users who are comfortable to do these steps using the documentation provided in the SensiML Embedded SDK for how to interface their application with the generated ML libraries.

  4. Piccolo AI supports only a subset of target platform architectures including Intel x86, arm Cortex M/A microcontrollers, and generic gcc code compilers. SensiML Analytics Studio provides these as well as additional compilers and platforms.