> For the complete documentation index, see [llms.txt](https://docs.eyepop.ai/developer-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.eyepop.ai/developer-documentation/glossary/pop.md).

# Pop

A Pop is the core building block of EyePop.ai.

A Pop defines how visual data is processed by combining AI models, logic, prompts, and workflows into a reusable pipeline. Every image, video, livestream, or camera feed analyzed by EyePop runs through a Pop.

Think of a Pop as a visual intelligence workflow that transforms raw media into structured, actionable data.

### Why Pops Exist

Most real-world visual AI problems require more than a single model.

For example, reading a license plate may involve:

1. Detecting a vehicle
2. Locating the license plate
3. Cropping the plate
4. Running OCR
5. Returning structured results

A Pop allows these steps to be combined into a single reusable workflow.

### What Can a Pop Contain?

A Pop can include:

* Object detection models
* Classification models
* OCR models
* Tracking
* Segmentation
* Keypoint detection
* Vision Language Models (VLMs)
* Structured extraction
* Prompting logic
* Data transformations
* Multi-stage processing pipelines

### Examples

#### License Plate Reading

```
Vehicle Detection
    ↓
License Plate Detection
    ↓
OCR
    ↓
Structured Results
```

#### Construction Site Monitoring

```
Worker Detection
    ↓
PPE Analysis
    ↓
Safety Violation Classification
```

#### Retail Shelf Analytics

```
Product Detection
    ↓
Product Classification
    ↓
Inventory Reporting
```

#### Visual Intelligence

```
Object Detection
    ↓
Crop Object
    ↓
Vision Language Model
    ↓
Structured Insights
```

### Inputs

A Pop can process:

* Images
* Video files
* Livestreams
* RTSP camera feeds
* RTMP streams
* WebRTC streams

### Outputs

A Pop returns structured JSON data that can be consumed by applications, dashboards, workflows, or business systems.

Outputs may include:

* Bounding boxes
* Labels
* Counts
* OCR results
* Tracking IDs
* Classifications
* Structured fields
* Event detections
* Visual Intelligence responses

### Reusability

Once created, a Pop can be used across multiple environments:

* Dashboard applications
* REST APIs
* SDKs
* Mobile applications
* Livestreams
* On-prem deployments
* Permanent Sessions

This allows teams to build a workflow once and deploy it anywhere.

### Deployment Options

A Pop can run in:

#### Cloud

Hosted and managed by EyePop.ai.

#### On-Prem

Deployed within a customer’s infrastructure.

#### Hybrid

A combination of cloud and edge processing.

#### Permanent Session

A dedicated runtime where a single Pop remains continuously active for low-latency and real-time workloads.

### Best Practices

When designing a Pop:

* Start with the simplest workflow that solves the problem.
* Use detection models to narrow the scope before running more expensive analysis.
* Return structured outputs whenever possible.
* Test with representative production data.
* Reuse Pops across applications rather than duplicating workflows.