[3/3] Anomaly detection tool (crops dataset)

Anomaly detection tool for crops dataset automates the identification of anomalous crop images. By inputting any image URL, it generates embedding vectors and compares them against a comprehensive crop database. The tool determines if the image matches known crops or flags it as an anomaly, enhancing agricultural monitoring and decision-making.

7/8/2025
17 nodes
Complex
spmntyrle9ydvwfamanualcomplexsticky noteexecuteworkflowtriggeradvancedapiintegration
Categories:
Complex WorkflowManual TriggeredBusiness Process Automation
Integrations:
Sticky NoteExecuteWorkflowTrigger

Target Audience

Target Audience


- Agricultural Researchers: Individuals studying crop varieties and their anomalies.
- Farmers and Agriculturalists: Those who want to monitor their crops for any unusual characteristics.
- Data Scientists: Professionals interested in applying machine learning techniques to agricultural datasets.
- Developers: Individuals looking to integrate anomaly detection into agricultural applications.

Problem Solved

Problem Solved


This workflow addresses the challenge of identifying anomalous crops in a dataset by comparing input images against a collection of known crop images. It automates the detection process, providing a text message indicating whether the input image depicts a known crop or an anomaly, thus enhancing crop management and monitoring.

Workflow Steps

Workflow Steps


1. Execute Workflow Trigger: The workflow is initiated manually by providing an image URL.
2. Image URL Hardcode: The provided image URL is stored for further processing.
3. Variables for Medoids: Essential parameters for accessing the Qdrant collection are set, including the Qdrant Cloud URL and collection name.
4. Total Points in Collection: Retrieves the total number of crop images stored in the Qdrant collection.
5. Each Crop Counts: Counts how many different crop classes are present in the collection.
6. Embed Image: The input image is embedded using the Voyage AI API to create a vector representation.
7. Get Similarity of Medoids: The embedded image is compared to the stored crop images in Qdrant to find similar classes based on predefined thresholds.
8. Compare Scores: The scores of the crops are analyzed. If the input image scores below the thresholds for all known crops, it is flagged as an anomaly.
9. Output Result: A message is returned indicating whether the input image is similar to existing crops or if it represents an unknown anomaly.

Customization Guide

Customization Guide


- Image Source: Users can modify the image source by changing the input in the Execute Workflow Trigger node.
- Threshold Parameters: Adjust the threshold types in the Compare Scores node to fine-tune the sensitivity of anomaly detection.
- Crop Dataset: Users can replace the existing crop dataset in Qdrant with their own dataset by uploading a new collection.
- Embedding Model: Change the embedding model in the Embed Image node to use different models if necessary, depending on the requirements of the analysis.