This is an explanatory material of Actcast, an edge-computing platform service, which is being developed by Idein Inc.
Actcast links various events and data in physical world to the web through deep learning inference on the edge.
Training Recognition models Can obtain algorithms by learning data for various pattern recognition tasks. notebook notebook digital clock glasses light stand potted plant book
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
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?
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
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
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
• 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.
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
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