keerthikoganti/lipstick-image-dataset

Dataset Card for keerthikoganti/lipstick-image-dataset

<!-- Provide a quick summary of the dataset. -->

Dataset Details

Dataset Description

<!-- Provide a longer summary of what this dataset is. --> This dataset consists of images labeled as lipstick (1) or no_lipstick (0). It was created as part of a classroom exercise in supervised learning and data augmentation, with the goal of practicing binary image classification and experimenting with dataset curation, preprocessing, and augmentation.

• —Curated by: Fall 2025 24-679 course at Carnegie Mellon
• —Shared by : Keerthi Koganti
• —Language(s) (NLP): N/A (Image Dataset)
• —License: Carnegie Mellon

Uses

<!-- Address questions around how the dataset is intended to be used. -->

Direct Use

<!-- This section describes suitable use cases for the dataset. -->

Training and evaluating binary image classification models (lipstick vs. no lipstick).

Experimenting with image preprocessing techniques (resizing, normalization, augmentation).

Teaching end-to-end machine learning workflows: data collection, labeling, augmentation, model training, and evaluation.

This data can be used to organize makeup products. Check how many of each products you have in your collection. Help address overconsumption.

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->

Production deployment in cosmetics detection systems.

Any application in safety-critical or automated decision-making contexts like marketing

Generalization to diverse real-world datasets without additional, larger-scale data collection.

Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

The dataset includes two splits:

original: Manually collected images (raw photographs).

augmented: Expanded dataset using synthetic transformations to balance classes and increase robustness.

Each row includes:

image: an image file (224×224 px, JPEG/PNG).

label: class label (0 = no_lipstick, 1 = lipstick).

Dataset Creation

Curation Rationale

<!-- Motivation for the creation of this dataset. -->

The dataset was curated to provide a simple, conceptually clear classification problem for learning purposes. Lipstick/no-lipstick classification was chosen because it is visually distinguishable and easy to capture.

Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

Data Collection: Original images were collected by the dataset creator using personal photography of makeup products.

Labels: Assigned manually at the time of collection.

Augmentation: Additional images generated via rotations, flips, cropping, brightness/contrast adjustments, and other transformations.

Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->

Original data: Keerthi Koganti Augmented data: Generated by using standard augmentation tools.

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

Small sample size: Limited number of original images, restricting model generalization.

Synthetic augmentation: Augmented data may not fully reflect natural real-world variation in lighting, backgrounds, or product appearances.

Domain bias: Images come from a single source specific lipstick product, background, and setting, not representative of all makeup products globally.

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Use primarily for teaching and demonstration of binary image classification.

Do not generalize conclusions beyond this dataset.

Highlight dataset limitations when presenting results, especially regarding data bias, augmentation, and generalization to real-world conditions.

Dataset Card Contact

Keerthi Koganti (Carnegie Mellon University) — kkoganti@andrew.cmu.edu