Start
AI Anomaly Detection
for Free

The AI model automatically learns normal behavior from diverse sensor data.
When signs of anomalies are detected, it immediately raises an alert.

Try Beta Version
...

Easy to Use

No knowledge of AI is required. Anyone can create AI models simply by uploading data.

Free Model Creation

Creating models and calculating scores can be done for free.
You can choose a paid plan after confirming the effectiveness.

Rich Features

We offer a variety of features according to each pricing plan,
including email notifications, data uploads via API, and management by multiple members.

Start anomaly detection
in 3 steps:

- Upload training data of normal operations and create an AI model.

- Upload test data to check for anomalies.

- Review the anomaly scores calculated by the AI on the screen. If the scores exceed the threshold, inspect the equipment.

...

Let's start for free!

Try Beta Version

Prices*

FAQ

We support two types of CSV files:
- Wide-format CSV files, where each row represents sensor values at each timestamp.
- Long-format CSV files, containing three columns: "Timestamp," "Sensor Name," and "Sensor Value."
For both data types, the first row is treated as the header row. In the case of long-format files, you need to specify the column numbers corresponding to "Timestamp," "Sensor Name," and "Sensor Value"; other columns will be ignored.

Sample Wide-format CSV
Sample Long-format CSV

You can upload CSV files by selecting them on the service interface or by dragging and dropping them onto the screen. Additionally, for users on the Light plan or higher, data can also be uploaded via API. This allows for automatic data uploads using scheduled jobs, such as cron jobs.

The AI model learns from the normal data uploaded in advance. The anomaly score becomes high when the test data shows behavior that deviates from the learned normal behavior. The accuracy of the AI model depends on the data used during training. Therefore, we recommend creating a model using both normal and abnormal data with the free plan first. This will help you determine whether the system can detect early signs of anomalies and whether the rate of false positives is acceptable before transitioning to a paid plan.

The anomaly score is calculated on an hourly basis. If the intervals between timestamps in the sensor data are shorter than one hour, the data will be averaged over each hour before calculating the anomaly score. If you need to detect early signs of anomalies quickly, we recommend uploading data every hour using the API or other means.

No, any numerical data in CSV format can be used to create models for anomaly detection. For example, you can convert the output of the sar command or web server access logs into CSV format and use them for anomaly detection.

We support credit card payments through Stripe.

For detailed instructions, please refer to the User Manual