Crop Anomaly Detection Workflow

This n8n workflow is designed for detecting anomalies in crop images by comparing them against a pre-established dataset stored in Qdrant. It leverages AI embeddings and clustering to identify whether a new image aligns with known crop classes or represents an anomaly. The process begins with embedding an input image via Voyage AI API, then querying Qdrant for similarity scores against crop clusters. Based on threshold comparisons, it determines if the image belongs to an existing crop class or if it is anomalous. This setup is ideal for precision agriculture, monitoring crop health, and identifying unusual crop phenomena.

**Workflow steps include:**

1. Triggered by an image URL input.

2. Embedding the image using Voyage AI’s multimodal model.

3. Querying Qdrant’s collection for similarity to crop medoids and cluster centers.

4. Comparing similarity scores against predefined thresholds.

5. Generating a descriptive message indicating whether the crop is recognized or anomalous.

**Key nodes and actions:**

– ApiRequest to Voyage AI for embedding.

– Qdrant API requests for similarity and clustering data.

– Python code to evaluate scores and detect anomalies.

– Sticky notes for guidance and descriptions.

This workflow is valuable for farms or agricultural research where rapid, automated crop health and anomaly assessments are essential.