arjunvettickal/RoboXS

---
pretty_name: RoboXS
task_categories:
- object-detection
tags:
- yolo
- computer-vision
- object-detection
license: mit
---

# RoboXS- Robotics Extended test Set

## Overview
This repository contains 4 separate YOLO-format batches:
A total of 21 parts of Robotics dataset were available to prepare the custom test set used to specifically assess the robustness of object detectors. The proposed idea was to create a test set of 80 images
across 10 object classes, split into four increasing-scene-difficulty equal batches
- `b1`
- `b2`
- `b3`
- `b4`

Each batch contains:
- `data.yaml`
- `images/test/*.jpg`
- `labels/test/*.txt`

## Format
This dataset is stored in raw YOLO format.

Each annotation row is:
`class x_center y_center width height`

Coordinates are normalized.

## Batch structure

### b1
- images: `b1/images/test/`
- labels: `b1/labels/test/`
- classes: see `b1/data.yaml`

### b2
- images: `b2/images/test/`
- labels: `b2/labels/test/`
- classes: see `b2/data.yaml`

### b3
- images: `b3/images/test/`
- labels: `b3/labels/test/`
- classes: see `b3/data.yaml`

### b4
- images: `b4/images/test/`
- labels: `b4/labels/test/`
- classes: see `b4/data.yaml`

## Notes
The batches are preserved exactly as exported.
This repository is intended for download and use in YOLO workflows.