README.md

# Real Lifebuoy Dataset for Object Detection

## Overview
This dataset contains images of real lifebuoy for object detection tasks. It can be used to train and evaluate object detection models.

Demo Example 1:

[![Lifebuoy Detection Video](https://img.youtube.com/vi/43IF39-LySI/0.jpg)](https://www.youtube.com/watch?v=43IF39-LySI)

The video is available on video/detr_finetuned_2.mp4 or by clicking the image youtube link.

Demo Example 2:

Left hand side: 1th finetuned with virtual dataset
Right hand side: 2nd further finetuned with real dataset

[![Lifebuoy Detection Video](https://img.youtube.com/vi/Vyb_iCp4bes/0.jpg)](https://www.youtube.com/watch?v=Vyb_iCp4bes)

The video is available on video/detr_finetuned_1vs2.mp4 or by clicking the image youtube link.

## Dataset Structure

### Data Instances

A data point comprises an image and its object annotations.

```
{
  'image_id': 1, 
  'image': <PIL.PngImagePlugin.PngImageFile image mode=RGB size=640x480 at 0x7781535B67C0>
  'image_path': 'images/2024-10-23-20-57-55_mid_305.png', 
  'width': 640, 
  'height': 480, 
  'objects': 
    {
      'id': [1], 
      'area': [4110.0], 
      'bbox': [[262.0, 312.0, 137.0, 30.0]], 
      'category': [0]
    }, 
}

```

### Data Fields

- `image_id`: the image id
- `width`: the image width
- `height`: the image height
- `objects`: a dictionary containing bounding box metadata for the objects present on the image
  - `id`: the annotation id
  - `area`: the area of the bounding box
  - `bbox`: the object's bounding box (in the [coco](https://albumentations.ai/docs/getting_started/bounding_boxes_augmentation/#coco) format)
  - `category`: the object's category, with possible values including 
    - `Lifebuoy` (0)

  
### Data Splits

- `Training dataset` (540)
    - `Real`
        - `Lifebuoy` (540)

- `Val dataset` (135)
    - `Real`
        - `Lifebuoy` (135)


## Usage
```
from datasets import load_dataset

dataset = load_dataset("ARG-NCTU/Real_Lifebuoy_dataset_2024")
```