cassieli226/lipproducts-dataset

Dataset Card for Lip Products Dataset

Dataset Description

This dataset was collected to study relationships between physical characteristics of lip products (tube size, price, color) and their product weight (grams).
It was created as part of a coursework assignment on tabular dataset building and augmentation.

• Repository: cassieli226/lipproducts-dataset
• Author: cassieli226
• License: CC BY-NC-SA 4.0
• Intended Use: Academic / coursework

Dataset Summary

• Small tabular dataset of lip products with physical attributes and weight.
• Regression target: Grams (net product weight).
• Includes manually measured samples and synthetic augmentations.

Supported Tasks

• Regression: Predicting product weight (grams) from physical and retail features.

Dataset Structure

Data Instances

Each row represents one lip product with its features and target weight.

Data Fields

• RGB (string): Average color of the lip product in "R,G,B" format
• Diameter_mm (float): Diameter of the product tube in millimeters
• Length_mm (float): Length of the product tube in millimeters
• Price (float): Retail price in USD
• Grams (float): Net product weight in grams (target variable)

Data Splits

• original: 30 manually measured samples
• augmented: 350 synthetic samples (via augmentation)

Dataset Creation

Source Data

• Collected from a personal collection of lip products.
• Measurements obtained with a ruler and product descriptions.
• RGB values extracted from swatch photos and averaged.

Preprocessing and Augmentation

Two augmentation strategies were applied:

1. Numeric Jitter

   • Gaussian noise added to numeric features and RGB channels (±5).
   • Simulates natural measurement and shade variability.

2. Mixup

   • Convex combinations of pairs of rows.
   • Applied to numeric features and RGB values.
   • Grams clipped to maximum physical capacity:
     ( \text{max grams} \approx \pi \times (Diameter/2)^2 \times Length / 1000 )

Considerations for Using the Data

Limitations

• Small initial sample size (30).
• Prices vary regionally and are not inflation-adjusted.
• RGB approximates perceived color and may differ under different lighting.

Ethical Notes

• Dataset is non-sensitive, collected from personal items.
• No private or personally identifiable information is included.
• Augmentation constrained by physical plausibility.

Additional Information

Dataset Curators

Prepared by cassieli226 for coursework.

Licensing Information

CC BY-NC-SA 4.0
Free for academic and research use, not for commercial purposes.

Contributions

ChatGPT was used to assist with preprocessing, augmentation code (numeric jitter & mixup), and debugging.
All augmentation parameters and physical constraints were designed by the dataset creator.

Example Usage

from datasets import load_dataset

ds = load_dataset("cassieli226/lipproducts-dataset")
print(ds)
print(ds["original"].to_pandas().head())

##Contact: [email protected]