Diverse Dataset for Image File Fragment Classification
Author Information
Author(s): Tavassoli Behnam, Naghshbandi Zhino, Teimouri Mehdi
Primary Institution: University of Tehran
Hypothesis
A diverse dataset is crucial for effectively training and evaluating models in image file fragment classification.
Conclusion
The dataset includes 501,000 image file fragments from various formats and content types, facilitating better evaluation of classification methods.
Supporting Evidence
- The dataset includes fragments from five common image formats: JPG, BMP, GIF, PNG, and TIFF.
- Images were categorized into three content types: Nature, People, and Medical.
- Each image was processed with various settings to ensure diversity in the dataset.
Takeaway
This study created a large collection of image pieces to help computers learn how to recognize different types of images better.
Methodology
The dataset was created by collecting images from the internet, categorizing them, and processing them into fragments using various settings.
Potential Biases
The use of specific image formats and limited content categories may introduce bias in the dataset's applicability.
Limitations
The study only focused on five common image formats and three content categories, which may not represent all real-world scenarios.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website