Codatta/Appliance-Knobs

---
license: openrail
tags:
- computer-vision
- image-classification
- object-detection
- 3d-understanding
- industrial-design
- Robotics
pretty_name: Appliance Knobs
dataset_info:
features:
- name: id
dtype: string
- name: image1
dtype: image
- name: image2
dtype: image
splits:
- name: train
num_bytes: 1741995315
num_examples: 408
download_size: 1712628607
dataset_size: 1741995315
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
language:
- en
size_categories:
- n<1K
---

# Appliance Knobs

## Overview

**Appliance Knobs** is a high-resolution dataset curated by **Codatta**, designed to support fine-grained object understanding, 3D shape estimation, and state recognition tasks.

This collection focuses on electrical appliance knobs and rotary controls. Its defining feature is the **paired image set** structure: every data entry captures the same specific knob from two distinct and correlated angles:
* **Front View:** A direct shot showing indicators and markings.
* **Side View:** A profile shot showing depth, height, and texture.

The dataset is filtered to ensure high fidelity, making it suitable for industrial design analysis, robotics, and generative AI applications requiring detailed reference material.

## Dataset Contents

Each entry in the dataset consists of a unique identifier and two high-quality images.

### Data Fields
* **`id`** (string): Unique identifier for the knob/appliance sample.
* **`image1`** (image): **Front View**. A direct frontal shot of the knob, clearly showing the face, markings, and position indicators.
* **`image2`** (image): **Side View**. A profile or oblique angle shot of the same knob to showcase its height, depth, material texture, and grip patterns.

### Quality Standards
* **Clear & Unoccluded:** All images have been manually verified to ensure the knob is the primary focus, free from obstruction by hands, wires, or other objects.
* **Lighting:** Consistent lighting is used to highlight the texture and markings of the controls.

## Key Statistics

* **Total Examples:** 408 paired samples.
* **Dataset Size:** ~1.74 GB (indicating high-resolution imagery).
* **Views per Sample:** 2 (Front and Side).
* **Language:** English (`en`).

## Usage

This dataset is optimized for tasks that benefit from multi-view correlation and high-resolution texture details.

**Supported Tasks:**
* **Multi-View Object Recognition:** Identifying objects using correlated information from different viewpoints.
* **3D Shape Reconstruction:** Inferring the 3D structure and depth of knobs based on the front and side profiles.
* **Knob State/Angle Estimation:** Training models to read the precise setting or angle of a dial.
* **Generative AI Training:** Serving as high-quality reference data for training LoRAs or ControlNets for specific industrial components.

**Python Usage Example:**
You can load and visualize the paired images side-by-side using the following code:

```python
from datasets import load_dataset
import matplotlib.pyplot as plt

# Load the dataset
ds = load_dataset("Codatta/appliance-knobs-dual-view", split="train")

# Get a sample
sample = ds[0]

# Visualize Front vs Side view
fig, axes = plt.subplots(1, 2, figsize=(10, 5))

axes[0].imshow(sample['image1'])
axes[0].set_title("Front View (Image 1)")
axes[0].axis('off')

axes[1].imshow(sample['image2'])
axes[1].set_title("Side View (Image 2)")
axes[1].axis('off')

plt.show()
```

## License and Open-Source Details
* **License:** This dataset is released under the **OpenRAIL** license.