KYOCERA enhances its AI-based Online Image Classifying Service “Labellio” with Higher Capacity for Big Data
Service allows users to easily create image recognition models with new cost-effective options; no expertise in deep learning or programming required
24 October 2017
Kyoto/London – Kyocera Corporation (President: Hideo Tanimoto) announced today that its subsidiary, Kyocera Communication Systems Co., Ltd. (hereafter “KCCS”), will launch three new fee-based plans for Labellio, its online image classifying service which enables users to quickly and easily classify digital images by using Artificial Intelligence (AI)-powered deep learning technology — without the need for programming or deep learning expertise. The new service plans will be available on October 30.
Today’s image recognition technologies are almost as accurate as the human eye. However, most of these systems require user expertise in AI, experience in deep learning software, and/or costly computing equipment. To help democratize this powerful technology, KCCS introduced its free Labellio service in January 2016 as a simple image classification tool — where users can simply upload images to the cloud via a PC or smartphone from virtually anywhere, eliminating the need for programming or deep learning expertise.
The three new fee-based Labellio plans incorporate feedback from users worldwide who have created thousands of image recognition models. As a result, the new Labellio services enable highly accurate image recognition models with online technical support at a low cost.
Users will be able to choose among four plans, offering a wide range of usage volumes and algorithms for companies interested in verifying advanced technologies or applying image recognition technology to commercial products or services:
- Existing free plan mainly for beginners.
- Light: for those users who would like to try the latest algorithms such as Inception and ResNet.
- Standard: for those users who would like to train their model for a few days to increase accuracy.
- Premium: for those users who do not want to have a limit on the monthly number of models created.
Labellio Image Classification Process
Use Case Example: Automation of manufacturer’s inspection process using image recognition
Main features
1. Create image recognition models easily at a low cost
At a monthly fee starting from USD10, Labellio lets users easily create image recognition models without the need for deep learning or programming expertise. The model can be used to classify new images using a PC or smartphone.
2. Adjust deep-learning parameters to improve accuracy
By adjusting deep learning parameters, including algorithm, batch size or learning rate, Labellio enables users to do trial and error to improve the model accuracy.
3. Download the created model
The image recognition models created by users can be downloaded* to incorporate into commercial products, services and applications.
* Image recognition models are only downloadable with Standard and Premium plans
4. Support from AI specialists
Users can consult KCCS’ AI support team by messaging them whenever they want to improve accuracy or seek technical consultation. The team will respond within one business day on a best-efforts basis.
Service Overview
| Plan | Free | Light | Standard | Premium |
|---|---|---|---|---|
| Monthly fee (USD) | Free | 10 | 300 | 1,500 |
| Number of models created / month | 10 | 10 | 30 | limitless |
| Number of models saved | 5 | 20 | 50 | 100 |
| Images stored | 1GB | 10GB | 50GB | 100GB |
| Max iteration / model | 1,000 | 1,000 | 10,000 | 500,000 |
| Algorithm | MobileNet | MobileNet Inception ResNet | MobileNet Inception ResNet | MobileNet Inception ResNet |
| Model downloads / month | 0 | 0 | 100 | 100 |
| API usage / month | 1,000 | 5,000 | 10,000 | 100,000 |
| Website | https://www.labell.io | |||
Note: All company and product names listed herein are trademarks or registered trademarks of the respective holders.
For more information on Kyocera: www.kyocera.eu
Editorial Material
- Press information (pdf) (539.10 KB)
Contact for technical information:
Mr. Ryo Fujita
E-mail: labellio@kccs.co.jp