README.md

---
license: cc-by-nc-2.0
task_categories:
- object-detection
language:
- en
tags:
- cellphone
- mobile phone
- phone
- cell phone
pretty_name: Cellphones in the Wild
---
# Dataset Card for Cellphones in the Wild

CITW is a small dataset that contains bounding box annotations of cellphones in images.  

## Dataset Details

### Dataset Description

CITW (Cellphones in the Wild) is a collection of images that contain one or more cell
phones in them, along with their corresponding bounding box annotations. CITW was distiled
from COCO 2017, where only the images and annotations containing a cellphone were kept. The
structure and annotations were adapted from COCO to be Huggingface compatible. 


- **Curated by:** Michael Grüner <[[email protected]](mailto:[email protected])>
- **Funded by:** [RidgeRun.ai](www.ridgerun.ai)
- **License:** CC-BY-NC-2.0

### Dataset Sources

- **Repository:** [https://huggingface.co/datasets/ridgerun-ai/citw-v0.1](https://huggingface.co/datasets/ridgerun-ai/citw-v0.1)
- **Demo:** TBD

## Uses

### Direct Use

CITW is meant to be used to train cellphone detectors.

### Out-of-Scope Use

The dataset only contains samples of mobile phones, and will not work for other types of phones, like office, faxes, or public phones.

## Dataset Structure

The dataset can be found within the `data` directory. It contains two splits: `train` and `val`, which are represented as subdirectories.
Within each split, you'll find the images in JPEG format, as well as a `metadata.jsonl` file.

The `metadata.jsonl` contains one entry per line. Each entry represents an image. The annotations can be found under the `objects` object.
This object contains a list of bounding boxes (which itself is a list), and a list of categories (which there is only one: 0).

A single bounding box is annotated as: `[x, y, width, height]`.

There is a single category: 0, which corresponds evidently to the cellphone class. 

### Entry Example

An example of a single entry is:
```
{
  "file_name": "000000253967.jpg",
  "objects": {
    "bbox": [
      [16.31, 104.46, 33.54, 43.17],
      [277.55, 146.1, 17.99, 58.69],
      [436.56, 130.99, 23.33, 42.09]
    ],
    "categories": [0, 0, 0]
  }
}
```

In this example, the image contains 3 cellphones. This is reflected in the 3 bounding boxes and the list of 3 cellphone categories.

## Dataset Creation

### Source Data

#### Data Collection and Processing

The dataset is a distillation of COCO2017. The repository contains a `coco2citw.py` script that automates this process.

#### Who are the source data producers?

Plase refer to the [COCO Challenge homepage](https://cocodataset.org/#home) for information of the original production process.


#### Personal and Sensitive Information

To the best of our knowledge, there are no personal and sensitive information in this dataset.

## Bias, Risks, and Limitations

This dataset is limited to cellphone models that were produced before 2017. Similar objects (like walkie-talkies, 
portable game consoles or calculatores) may be eventually confused.


## Glossary

CITW: Cellphones in the Wild

## Dataset Card Authors

Michael Grüner <[[email protected]](mailto:[email protected])>

## Dataset Card Contact

RidgeRun.ai <[[email protected]](mailto:[email protected])>