Hanwha Vision WiseDetector

What is WiseDetector?

The Wise Detector is a Machine Learning Model (ML) where the user can train the camera to detect a user-specified object. It consists of two main components: The Wisenet Device Manager and the WiseAl Application.

Key Features:

• User-defined object detection
• Easy to set-up
• Supported by P series AI cameras
• License Free

Components

1. The Wisenet Device Manager:

   A free Hanwha PC application that helps scan and configure Hanwha's IP devices. It allows users to train the WiseDetector Model to detect a specific object using the Wise Detector option.

2. The WiseAl Application:

   An on-camera application that contains AI detection and video analytics. The trained WiseDetector AI model runs within this application on Hanwha P series IP cameras.

The process involves using the Wisenet Device Manager on a Windows PC to select training samples, train the model, and then upload it to the WiseAl Application on a compatible camera.

Simple and Easy to Use: Object Training

The training process is initiated within the Wisenet Device Manager. Users select sample images of the object of interest from the target camera and draw bounding boxes around it. The system allows for iterative training until satisfactory results are achieved.

Training Steps:

1. In the Wisenet Device Manager, right-click on the target camera and select 'Wise Detector'.
2. Create a new object with a name that does not include spaces or special characters.
3. Click 'Snapshot' when the target object is visible and draw a bounding box around it.
4. Ensure the bounding box accurately contains the object.
5. Click the 'Train' button.
6. Repeat these steps until the results are satisfactory.

The Wise Detector Model provides feedback on training results, allowing users to observe the progress before uploading the model to the camera.

Color Filtering and Sensitivity Level (Optional)

Color Filtering: WiseDetector can optionally filter detections based on up to two colors. This feature is not available in night mode when the camera generates a black and white image.

Sensitivity: Users can adjust the sensitivity setting to increase or decrease the accuracy threshold of the model's detection.

Uploading Data into P Series AI Cameras

Once trained and configured, the WiseDetector model can be uploaded to the WiseAl application on compatible cameras. Up to two models can be loaded onto a single camera and enabled or disabled as needed. Detections and metadata can trigger video analytics rules like line crossing or loitering.

P Series AI Cameras with WiseDetector

WiseDetector is compatible with P series AI cameras running the WiseAl application. It works alongside the existing AI detection features of these cameras.

Compatible Models:

• PNB-A6001, PNB-A9001, PNB-A6081R, PNB-A9081R, PNV-A6081R, PNV-A9081R, PND-A6081RV, PND-A9081RV, PND-A6081RF, PND-A9081RF

Required Firmware and App Versions:

• For PNB-A6001, PNB-A9001, PNB-A6081R, PNB-A9081R, PNV-A6081R, PNV-A9081R, PND-A6081RV, PND-A9081RV, PND-A6081RF, PND-A9081RF: WiseAl app version 1.01.01 or higher; firmware version 2.21.05 or higher (bridge f/w version: 2.19.99).
• For PNB-A6001, PNB-A9001, PNB-A6081R, PNB-A9081R, PNV-A6081R, PNV-A9081R, PND-A6081RV, PND-A9081RV, PND-A6081RF, PND-A9081RF: WiseAl app version 1.01.01 or higher; firmware version 2.21.12 or higher (bridge f/w version: 2.19.99).

Guidelines for Accurate Detection

• The user-specific object should maintain a consistent shape and size for successful detection. Variations may lead to inaccurate results.
• Changes in object size or occlusion can impact detection accuracy. Ensure sample images are representative of the actual environment, including lighting conditions (day, night, shade), weather (rain, snow), and background scenery.
• The target camera must be compatible, have the correct firmware installed, and be positioned correctly with the Field of View (FoV) configured. Changes to camera position or FoV after model upload can cause inaccuracies.
• Up to two separate Wise Detectors can be uploaded to P Series AI cameras.
• A trained WiseDetector model can be reused for other cameras, but new training samples specific to those cameras must be included during the training process.

Contact Information

Head Office: 6, Pangyo-ro 319beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, 13488 South KOREA. Tel: 1588-5772

For more information, please visit hanwhavision.com.