sdiaeyu6n/tu-berlin

Dataset Card for TU Berline Dataset

This dataset card aims to provide comprehensive information about the TU Berlin dataset, a collection of hand-drawn sketches used for training and evaluating sketch classification models.

Dataset Details

Dataset Description

The TU Berlin dataset is a large-scale collection of hand-drawn sketches curated by the research team at TU Berlin. The dataset includes 20,000 unique sketches across 250 object categories, contributed by participants from around the world. The primary purpose of this dataset is to facilitate research in the field of computer vision, particularly for tasks related to sketch recognition and classification.

• Curated by: TU Berlin research team
• Shared by [optional]: TU Berlin

Dataset Sources

• Source: TU Berlin Dataset Source
• Paper: TU Berlin Dataset Paper

Uses

Direct Use

The dataset is intended for use in developing and evaluating sketch recognition algorithms. It is suitable for tasks such as:

• Training sketch classification models
• Evaluating the performance of sketch recognition systems
• Conducting research in computer vision and machine learning related to hand-drawn images

Out-of-Scope Use

The dataset is not suitable for use cases that require high-resolution images or photographs. It is also not intended for tasks unrelated to sketch recognition, such as natural image classification.

Dataset Structure

The dataset is organized into categories, each containing a collection of hand-drawn sketches. Each sketch is a black-and-white image representing an object from one of the predefined categories.

• Number of Categories: 250
• Number of Sketches: 20,000

Dataset Splits

I downloaded the TU Berlin dataset and split it into train set, validation set, and test set.

• Train Set:
  • Number of Examples: 16,000
  • Size: 480,609,419 bytes
• Validation Set:
  • Number of Examples: 2,000
  • Size: 59,693,656 bytes
• Test Set:
  • Number of Examples: 2,000
  • Size: 60,354,461 bytes
• Download Size: 589,085,954 bytes
• Total Dataset Size: 600,657,536 bytes

The data was split using the following code:

from sklearn.model_selection import train_test_split

train_data, temp_data = train_test_split(metadata, test_size=0.2, random_state=42)
val_data, test_data = train_test_split(temp_data, test_size=0.5, random_state=42)

Citation

BibTeX:

@article{eitz2012hdhso,
author={Eitz, Mathias and Hays, James and Alexa, Marc},
title={How Do Humans Sketch Objects?},
journal={ACM Trans. Graph. (Proc. SIGGRAPH)},
year={2012},
volume={31},
number={4},
pages = {44:1--44:10}
}