Point-Count App
  • 30 Apr 2025
  • 1 Minute to read
  • Dark
    Light

Point-Count App

  • Dark
    Light

Article summary

The Cogniac Point Count Application enables automated, AI-driven object counting within images and videos. It is ideal for use cases where you need to count distinct instances of a particular subject (e.g., people, cars, parts) with precision and scalability.


What Is the Point Count App?

The Point Count App is a specialized application within the Cogniac platform that:

  • Detects and counts visible instances of a specific object in visual media.

  • Generates accurate count labels that can be used for reporting, quality assurance, or triggering downstream workflows.

  • Learns from human feedback to continuously improve performance.


Application Configuration

Step 1: Create a New Count Application

  1. Log into your Cogniac account.

  2. Go to Application Builder.

  3. Select “Point Count” as the application type.

  4. Name your application (e.g., Inventory Shelf Counter, Worker Count Line 2).

Step 2: Set Subjects

  • Input Subject: Defines the incoming media to analyze.

  • Output Subject: Stores the processed results (typically the same as the input subject).

Step 3: Set Parameters

  • min_count: Smallest number of objects expected in the image (default is 0).

  • max_count: Largest number of expected objects (helps optimize detection and training performance).


Uploading and Processing Media

  1. Upload images or videos to the input subject.

  2. The application analyzes each media item and returns:

    • An object count

    • Visual overlays of counted instances

    • Confidence scores (optional)


Providing Feedback (Essential for Training)

Cogniac’s Point Count model is refined through user feedback.

How to Provide Feedback:

  • Open a processed image in the UI.

  • Confirm or correct the point count value shown.

  • The corrected count is stored in the app_data field with:

    • app_data_type: "count"

    • value: An integer representing the correct count

As you submit feedback:

  • The model re-trains periodically.

  • It automatically adjusts the min_count and max_count values if needed.


Output & Insights

The Point Count App outputs:

  • Count values for each image

  • Summary analytics via dashboards (optional)

  • Count-based event triggers via API (e.g., send alert if count > 10)

All outputs are stored in the output subject and accessible through the Cogniac API or UI.


Data Privacy & Model Control

  • All data is processed securely within the Cogniac platform.

  • You maintain full control over your application, subjects, and training data.

  • Access is role-based for added security.


Best Practices

  • Use high-resolution images for better accuracy.

  • Regularly review and correct results, especially early in deployment.

  • Choose min_count and max_count wisely to set expectations.

  • Keep subject naming consistent and descriptive.


Was this article helpful?

ESC

Eddy AI, facilitating knowledge discovery through conversational intelligence