28x28 QuickDraw sketches provide data for binary classification of flower versus not-flower images. The dataset is designed for training convolutional neural networks for computer vision tasks. Its origin and creation date are unknown.
Use Cases
- Train a CNN to classify images as 'flower' or 'not-flower' using the 28x28 pixel sketch data.
- Benchmark binary classification model performance on standardized, hand-drawn image data.
- Fine-tune pre-trained vision models on a stylized, sketch-based dataset for transfer learning.
- Develop data augmentation strategies specific to 28x28 grayscale hand-drawn sketches.
Strengths
- All images are uniformly sized at 28x28 pixels, simplifying model input pipelines.
- Data is curated for a clear binary classification task between flower and not-flower categories.
Limitations
- The total number of image samples (rows) is unknown, preventing assessment of dataset scale.
- The class balance between 'flower' and 'not-flower' sketches is unspecified, risking potential bias.
Provenance
- Source
- null
- Collection Method
- Derived from the QuickDraw dataset of hand-drawn sketches.
- Time Range
- null
- Freshness
- null
- Geography
- null