Documentation
AI'GORITHM
Algorithm examples

Algorithm examples

1. Monitoring candy bar movement on a conveyor belt

Imagine a scenario where we need to monitor candy bars moving along a conveyor belt. The goal is to determine if a candy bar moves entirely within a 3-second window or surpasses a defined size threshold more than 50 times. If either condition is met, notifications are sent via SMS. This process can be automated with an algorithm as follows:

Algorithm structure

1. Add a Model block to perform inference:

  • Click Add model and select an object detection model capable of identifying candy bars,

  • Ensure the model has been properly trained and includes the relevant categories.

2. Add a Session block:

  • Connect the Session block to the Model block. The Session defines a loop that takes a frame from the stream, applies the inferencing engine, and returns the predictions.

3. Add Prediction triggers:

  • Add a Speed trigger to identify candy bars moving entirely across the conveyor belt within 3 seconds,

  • Configure the following:

Categories: Select the relevant candy bar categories,

Seconds: Set to 3,

Distance: Set to 100%,

  • Add a Size trigger to identify candy bars exceeding the specified size thresholds,

  • Configure the following:

Categories: Select the relevant candy bar categories,

Operator: Set to “greater than”,

Height: Set to 80%,

Width: Set to 30%,

  • Connect both prediction triggers to the Session block.

4. Add a Prediction event:

  • Add a Sends sms action block to send real-time notifications to a specified phone number,

  • Connect the Speed trigger to the Send sms action block,

  • Configure the following:

Phone numbers list: Add the recipient’s phone number,

SMS body: Set the message (e.g., “The production line is moving too fast”).

5. Add an Aggregator block:

  • Connect the Aggregator block to the Size trigger block. This block groups (aggregates) events derived from prediction triggers.

6. Add an Aggregation trigger:

  • Add an Events quantity block to track when size violations occur more than 50 times for a single candy bar,

  • Configure the following:

Operator: Set to “greater than”

Number: Set to 50.

7. Add another Prediction event:

  • Add a Send sms action block to send real-time notifications to a specified phone number,

  • Connect the Events quantity block to Sends sms action block,

  • Configure the following:

Phone numbers list: Add the recipient’s phone number,

SMS body: Set the message (e.g., “Candy bars moving on the conveyor belt are too big.”).

2. Analyzing the amount of cards

Suppose we aim to determine the quantity of specific cards captured by both the top and right cameras. The algorithm might be structured as follows:

Algorithm structure

1. Add two Model blocks to perform inference:

  • Add one Model block for the top camera and another for the right camera,

  • Select object detection models capable of recognizing card symbols,

  • Ensure models have been properly trained and include the relevant categories.

2. Add two Session blocks:

  • Each Session block should be connected to one Model block. The Session defines a loop that takes a frame from the stream, applies the inferencing engine, and returns the predictions.

3. Add two Prediction triggers:

  • Add two Quantity triggers, one for the top camera and one for the right camera,

  • Configure the following:

Category: Select the relevant card symbols,

Operator: Set to “greater than”,

Number: Set to 1,

  • Each Quantity trigger block should be connected to one Session block.

4. Add an Aggregator block:

  • Connect the Aggregator block to the Quantity trigger blocks. This block groups (aggregates) events derived from prediction triggers.

5. Add an Aggregation trigger:

  • Add an Events quantity block to track when quantity violations occur more than 1 time,

  • Configure the following:

Operator: Set to “greater than”,

Number: Set to 1,

Enable Consider only events related to the same object.

6. Add a Prediction event:

  • Add a Make screenshot action block to to save an image of the detected cards,

  • Connect the Events quantity block to Make screenshot action block,

  • Configure the following:

Relative path: Add a relative path and use variables such as OccurredEventsNames and EventsDetailsJson to generate file paths dynamically.