List of datasets:
- Caltech-101 (subdataset containing only bikes, cars and faces)
- Digits (simple handwritten digits dataset with binary values)
- DigitsX (variation of Digits with multivalued pixels and slight gaussian blur)
- Four-by-four (artificial dataset for testing)
- Olivetti (AT&T Bell Labs faces dataset)
- Olivetti 3-bit (Olivetti dataset downscaled to 3-bit resolution)
- Olivetti Big (AT&T Bell Labs faces with original size)
- Olivetti Padded Uniform (Olivetti dataset with padded left and right borders so that width and height are divisible by four; paddings pixel values are uniformly distributed)
- Olivetti Small (Olivetti dataset smaller size)
- MNIST
- Self driving dataset (https://github.com/felipessalvatore/self_driving_data)
Most datasets have their original images split into its categories by a folder with its name indicating its class.
A compiled GoSPN dataset file in .data format is included
in every dataset for use in GoSPN.
For MNIST, we use GoMNIST to parse the dataset. We select only a portion of MNIST such that the selected digits are uniformly distributed between 0 and 9.