An Edge-Computing Platform

Pattern recognition with deep learning Lesser panda Input data Annotations Training Recognition models Can obtain algorithms by learning data for various pattern recognition tasks. notebook notebook digital clock glasses light stand potted plant book

Actcast is a platform service for IoT (Internet of Things) systems. It links various events and data in physical world to the web through deep learning inference on the edge. What is Actcast ｛・・・｝ raw data semantic data API call

Many possibilities with the combination of “Acts” and “Casts” Programs running on the edge to obtain real-world data Rules describing how to integrate data and web services Act Cast What kind of data do you collect? How do you use the data?

Use cases  Home security  In-store marketing  Acceptance system  Digital signage  Product inspection  Warehouse management  Cultivation management  Infrastructure monitoring and so forth Database Notification Analytics Visualization Detect events Classify objects Estimate status Detect anomalies Icons are designed by Freepik.com

Advantages of edge computing { “timestamp”: “xxxxxxxx”, “data”: [ {“age”: 26, “gender”: “female”}, {“age”: 33, “gender”: “male”}, {“age”: 19, “gender”: “female”}] } Illustrations are designed by Freepik.com Send raw data every frame Send semantic data only when it is necessary Compute on the server Compute on the edge ✓ Lower data transfer cost ✓ Lower server cost ✓ Lower privacy risk in comparison to computing on the server

How to use Actcast

You can use Raspberry Pi Our technology enables to use Raspberry Pi, a low price commodity computer, for executing deep learning inference on the edge. Watch our demo videos https://www.youtube.com/c/IdeinInc “Raspberry Pi” is a trademark of the Raspberry Pi Foundation. Raspberry Pi 3 $35 Raspberry Pi Zero $5 Compute Module 3 $30

Easy setup Things you need to prepare • Raspberry Pi • Camera Module • microSD card • AC adapter Write install and configuration files to microSD card Assemble and power on Raspberry Pi Your Raspberry Pi will appear on Actcast within minutes and be ready to use.

Hosting recognition models  Ready-to-use recognition models  Detection of faces, people, …  Age and gender estimation  Human pose estimation and so on  You can use your custom models  You can make a model public and share it with other users Custom models will be available after main release

Integration with web services  Re-format the data sent from devices according to destination service  Specify conditions on data to trigger transmission to web services

 Deployment of software to devices  Remote update of firmware  Device status monitoring Device management

$5～$50/month Cost comparison $5,000～$10,000/month $1~$2/1,000 frame Inference on a cloud service Case study example: Inference 2 images/sec ≈ 5M images/month Expected price

Roadmap  Late 2018 (α-release)  Free of charge  Up to 5 devices per user  Early 2019 (β-release, main release)  Plan for commercial use  SDK for user custom DNN model and apps  Workspaces  Unlimited number of devices  Support schedule is undecided, but we are working toward support of devices other than Raspberry Pi

Please use this contact only when you have interested in  Partnership with us related to Actcast  Media coverage and interviews @IdeinInc @IdeinInc Information Contact [email protected]