Cosine Similarity

Class: CosineSimilarityBlockV1

Source: inference.core.workflows.core_steps.math.cosine_similarity.v1.CosineSimilarityBlockV1

Calculate the cosine similarity between two embeddings.

A cosine similarity of 1 means the two embeddings are identical, while a cosine similarity of 0 means the two embeddings are orthogonal. Greater values indicate greater similarity.

Type identifier

Use the following identifier in step "type" field: roboflow_core/cosine_similarity@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Unique name of step in workflows.	❌

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections

Compatible Blocks

Check what blocks you can connect to Cosine Similarity in version v1.

• inputs: CLIP Embedding Model, Perception Encoder Embedding Model, Identify Changes
• outputs: SAM 3, OpenAI, Crop Visualization, PTZ Tracking (ONVIF).md), OpenAI, Google Gemini, Anthropic Claude, Label Visualization, Llama 3.2 Vision, Google Gemini, Seg Preview, Roboflow Dataset Upload, Dynamic Zone, Segment Anything 2 Model, Template Matching, Webhook Sink, Velocity, Distance Measurement, Camera Calibration, Line Counter Visualization, Keypoint Visualization

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds Cosine Similarity in version v1 has.

Bindings

• input

  • embedding_1 (embedding): Embedding 1.
  • embedding_2 (embedding): Embedding 2.

• output

  • similarity (float): Float value.

Example JSON definition of step Cosine Similarity in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/cosine_similarity@v1",
    "embedding_1": "$steps.clip_image.embedding",
    "embedding_2": "$steps.clip_text.embedding"
}