# Image Metadata

Metadata lets you attach custom key-value pairs to images in your Roboflow workspace. Use metadata to store structured information alongside your images — such as capture conditions, device identifiers, quality scores, or any domain-specific attributes — and then search, filter, and organize your data around those attributes.

## Overview

Each image can hold any number of metadata entries. An entry is a **key** (a name like `camera_id`) paired with a **value** (a string, number, or boolean).

| Value type | Examples                                    |
| ---------- | ------------------------------------------- |
| String     | `location: "warehouse-3"`, `shift: "night"` |
| Number     | `temperature: 72.5`, `quality_score: 95`    |
| Boolean    | `reviewed: true`, `is_night: false`         |

### Use cases

* **Capture context** — record camera ID, GPS coordinates, weather, lighting conditions
* **Quality tracking** — attach confidence scores, review status, annotator IDs
* **Data slicing** — filter your dataset by any attribute to build targeted training sets
* **External system linking** — store identifiers that connect images back to your internal tools

## Adding Metadata

You can add metadata to images through the web UI, the Python SDK, the REST API, or automatically via S3 Bucket Mirror.

### Web Application

{% stepper %}
{% step %}

#### Open an image

Open any image in your project.
{% endstep %}

{% step %}

#### Enter key and value

In the metadata section, enter a **key** in the first input and a **value** in the second input.
{% endstep %}

{% step %}

#### Add

Press **Enter** to save or click on Add
{% endstep %}
{% endstepper %}

Values are automatically parsed by type:

| Value entered    | Stored as                  |
| ---------------- | -------------------------- |
| `front`          | `"front"` (string)         |
| `95`             | `95` (number)              |
| `3.14`           | `3.14` (number)            |
| `true` / `false` | `true` / `false` (boolean) |

<figure><img src="/files/RfPha5Er2oycBtTLcQjn" alt=""><figcaption><p>Annotation Tool's metadata editor</p></figcaption></figure>

### Python SDK

Pass a `metadata` dictionary when uploading an image:

```python
import roboflow

rf = roboflow.Roboflow(api_key="YOUR_API_KEY")
project = rf.workspace("your-workspace").project("your-project")

project.upload(
    image_path="image.jpg",
    metadata={
        "camera_id": "cam001",
        "location": "warehouse-3",
        "temperature": 72.5,
        "is_night": False
    }
)
```

### REST API

#### Add metadata during upload

Include a `metadata` field (JSON-stringified) in the multipart form data when uploading an image:

```bash
curl -X POST "https://api.roboflow.com/dataset/your-dataset/upload?api_key=YOUR_API_KEY" \
  -F "name=image.jpg" \
  -F "split=train" \
  -F "file=@image.jpg" \
  -F 'metadata={"camera_id":"cam001","temperature":72.5}'
```

### S3 Bucket Mirror

When using [Bucket Mirror](/datasets/adding-data/upload-data-from-aws-gcp-and-azure/aws-s3-bucket.md) to sync images from an S3 bucket, you can attach metadata by placing a `.json` file alongside each image with the same base name:

```
my-bucket/
  images/
    photo_001.jpg
    photo_001.json      # metadata for photo_001.jpg
    photo_002.jpg
    photo_002.json      # metadata for photo_002.jpg
```

The JSON file contains your metadata as key-value pairs:

```json
{
    "camera_id": "cam001",
    "location": "warehouse-3",
    "capture": { "temperature": 72.5, "humidity": 45 }
}
```

**Nested objects are automatically flattened** using dot notation. The example above produces:

| Key                   | Value           |
| --------------------- | --------------- |
| `camera_id`           | `"cam001"`      |
| `location`            | `"warehouse-3"` |
| `capture.temperature` | `72.5`          |
| `capture.humidity`    | `45`            |

#### Metadata file constraints

* Maximum file size: **256 KB**
* Must be valid JSON
* `null` and `undefined` values are filtered out

#### Update strategies

Bucket Mirror supports different strategies for how synced metadata interacts with metadata you've set manually in the UI or via API:

| Strategy                        | Behavior                                                                     |
| ------------------------------- | ---------------------------------------------------------------------------- |
| **`mergeBucketWins`** (default) | Merges both sources. On key conflicts, the bucket value wins.                |
| **`mergeUserWins`**             | Merges both sources. On key conflicts, the user-set value wins.              |
| **`overwrite`**                 | Bucket metadata completely replaces all existing metadata.                   |
| **`untilFirstChange`**          | Syncs from bucket until a user manually edits metadata, then stops updating. |
| **`append`**                    | Only adds new keys from the bucket. Never overwrites existing keys.          |

## Searching by Metadata

Metadata is indexed and searchable in the [Asset Library](/workspaces/asset-library.md). Use the search bar to filter images by metadata values:

```
metadata.camera_id:"cam001"
metadata.quality_score>80
metadata.reviewed:true
```

You can combine metadata filters with other search filters:

```
metadata.location:"warehouse-3" AND class:forklift
```

The Asset Library also provides autocomplete for metadata keys and values based on what exists in your workspace.

## Key Naming Rules

Metadata keys must follow these rules:

| Rule                 | Detail                                                                 |
| -------------------- | ---------------------------------------------------------------------- |
| Allowed characters   | Letters (`a-z`, `A-Z`), numbers (`0-9`), underscores (`_`), dots (`.`) |
| First character      | Must be a letter, number, or underscore                                |
| Forbidden characters | Forward slashes (`/`) are not allowed                                  |

Valid keys: `camera_id`, `capture.temperature`, `_internal_ref`, `v2_score`

Invalid keys: `camera/id` (contains `/`), `.starts_with_dot` (starts with `.`), `has spaces` (contains spaces)

## Metadata vs. Tags

Both metadata and [tags](/datasets/manage-datasets/add-tags-to-images.md) help you organize images, but they serve different purposes:

|               | Tags                                 | Metadata                                   |
| ------------- | ------------------------------------ | ------------------------------------------ |
| **Structure** | Simple labels                        | Key-value pairs                            |
| **Values**    | No value, just a name                | String, number, or boolean                 |
| **Best for**  | Categorization, workflow status      | Structured attributes, measurements        |
| **Example**   | `reviewed`, `v2`, `needs-annotation` | `temperature: 72.5`, `camera_id: "cam001"` |

You can use both on the same image. For example, tag an image as `reviewed` and also store `reviewer: "alice"` and `confidence: 0.95` as metadata.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.roboflow.com/datasets/adding-data/image-metadata.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.