> ## Documentation Index
> Fetch the complete documentation index at: https://docs.worlds.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Training & Deployment

> Walk through the full flow from raw camera feeds to a trained model and automated workflow.

This guide walks through how to go from raw camera feeds to a working vision model and automated workflow. The process is iterative by design, expect to circle back and refine as you learn what your model does and doesn't catch.

## Before You Start

**Tip:** Define your use case before you touch anything. Know what events or objects you actually need to detect (a forklift entering a zone, inventory leaving storage, a pallet jack in use, etc). That goal shapes every decision below, from how you group cameras to what you tag.

**Warning:** If you're new to the platform, build and test in a non-production or demo environment first. Making changes directly in a live/production environment risks breaking real workflows other teams depend on.

## Step 1: Organize Cameras into Collections

A [**collection**](/collections-and-models/overview) is a folder that groups cameras which will share the same vision model.

**How to decide what goes together:**

* Group cameras that share visual context, similar environments, looking for similar things.
* Keep visually distinct feeds separate. An exterior camera and an interior camera will generally perform better as separate collections with separate models rather than one shared model.
* A camera isn't locked into one collection. The same camera can belong to multiple collections, for example, running a specialized model in one collection while also contributing to a general "people detection" collection.

**Example:** A two-camera collection covering a loading dock area, where both cameras track people, a forklift, inventory, and a pallet jack, just from different angles. Because both cameras are watching the same kind of activity in the same environment, they share one model.

## Step 2: Configure Zones

Once cameras are organized into a collection, configure [**zones**](/sites-and-cameras/cameras-folders-zones) within each camera view.

* Zones define specific areas of interest inside the frame (a storage area, a dock door, etc).
* This is the information that later feeds into the [workflow builder's](/agents-and-workflows/overview) evaluation logic, it's how the system knows *where* to look for an event, not just *what* to look for.

**Tip:** Set up zones with your end goal in mind. If you want to know when inventory is picked up and moved, or when a pallet jack leaves its storage spot, that's exactly the kind of zone to configure here.

## Step 3: Tag and Review Detections

This is where you move into the **tagging interface** and its companion, the **inference viewer**.

* The inference viewer shows your video feed alongside a timeline of every object detected and tracked during that period.
* Each object type (forklift, inventory, pallet jack, person, etc) is a label you've trained into the model, shown as its own track on the timeline.
* A solid, continuous block on the timeline means the object was tracked without interruption. A broken-up track means detection was lost and picked back up.

**Warning:** Broken tracks aren't necessarily a sign of a bad model. Occlusion is a common cause, if a forklift drives in front of a tracked object, the system may drop that detection and pick it up as a *new* track once it's visible again. This is expected behavior to watch for, not automatically a flaw to fix.

## Step 4: Train and Iterate the Model

* Train a model version, then review its performance using the inference viewer.
* This process is intentionally iterative. Expect to be on v2, v3, and beyond as you refine things, it's normal, not a sign something went wrong the first time.

**Tip:** If you notice false positives or missed detections, the fix is often to add a new object class so the model has better context for what it's looking at, rather than assuming the model itself is broken.

## Step 5: Move to the Workflow Builder

With collections, zones, and a trained model in place, the final step is building automated workflows (in the [workflow builder](/agents-and-workflows/overview)) that act on the zone and tag data established in Steps 1-4.

**Tip:** Look at existing workflows and models your team has already built. They're a useful reference point for structure and strategy before you build your own from scratch.

## Quick Reference: The Flow

1. **Collections** - group cameras by shared visual context and goal
2. **Zones** - define the specific areas that matter within each camera
3. **Tagging / Inference Review** - label objects, review detection tracks
4. **Model Training** - train, evaluate, iterate
5. [**Workflow Builder**](/agents-and-workflows/overview) - automate based on zones and tags