GANji: A Framework for Introductory AI Image Generation

Chandon Hamel; Mike Busch

doi:10.48550/arxiv.2509.24128

Paper details

Research paper

Other Computational

GANji: A Framework for Introductory AI Image Generation

Chandon Hamel, Mike Busch

arXiv·2025·10.48550/arxiv.2509.24128·arXiv:2509.24128

Samples

Each sample groups how it was prepared, characterized, and measured. Click a card to expand.

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

18 properties

SimulatedJapanese Kanji character images datasetStudied Material

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

15 properties

SimulatedJapanese Kanji character images datasetStudied Material

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

2 properties

SimulatedJapanese Kanji character images datasetStudied Material

Materials

Japanese Kanji character images dataset

Studied Material

Also known as

Kanji DatasetKanji characters

Variables

channels= grayscale

font_style= Mincho

sample_count= 10314

image_resolution_px

GANji: A Framework for Introductory AI Image Generation | Matter42 Literature

Properties

Type	Label	Value	Material	Sample	Source
Count Predicted 2d Candidates	VAE latent space size	64	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	128	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	256	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	79.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	91.3FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	86.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2458MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2464MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2560MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	22.2M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	23.8M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	26.9M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	GAN FID	74.6FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.4FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	48.8min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	184.1min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	69.2min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	4190MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	DDPM linear noise schedule start	0.0001	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—
Other Computational	DDPM linear noise schedule end	0.02	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—

Properties

Type	Label	Value	Material	Sample	Source
Count Predicted 2d Candidates	VAE latent space size	64	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	128	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	256	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	79.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	91.3FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	86.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2458MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2464MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2560MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	22.2M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	23.8M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	26.9M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	GAN FID	74.6FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.4FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	48.8min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	184.1min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	69.2min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	4190MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	DDPM linear noise schedule start	0.0001	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—
Other Computational	DDPM linear noise schedule end	0.02	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—

Properties

Type	Label	Value	Material	Sample	Source
Count Predicted 2d Candidates	VAE latent space size	64	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	128	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	256	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	79.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	91.3FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	86.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2458MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2464MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2560MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	22.2M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	23.8M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	26.9M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	GAN FID	74.6FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.4FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	48.8min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	184.1min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	69.2min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	4190MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	DDPM linear noise schedule start	0.0001	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—
Other Computational	DDPM linear noise schedule end	0.02	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—

Properties

Type	Label	Value	Material	Sample	Source
Count Predicted 2d Candidates	VAE latent space size	64	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	128	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Count Predicted 2d Candidates	VAE latent space size	256	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	79.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	91.3FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE FID	86.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE training time	14.4min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE sampling time	4s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2458MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2464MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE VRAM usage	2560MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	22.2M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	23.8M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	VAE model parameters	26.9M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the VA	—
Other Computational	GAN FID	74.6FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.9FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN FID	78.4FID	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	48.8min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	184.1min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN training time	69.2min	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN sampling time	3s	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	4190MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN VRAM usage	6046MB	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	GAN model parameters	8.7M	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the GA	—
Other Computational	DDPM linear noise schedule start	0.0001	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—
Other Computational	DDPM linear noise schedule end	0.02	Japanese Kanji character images dataset	Benchmark dataset instance used to train and evaluate the DD	—

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

GANji: A Framework for Introductory AI Image Generation

Samples

Benchmark dataset instance used to train and evaluate the VAE model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the GAN model on upscaled Kanji character images.

Benchmark dataset instance used to train and evaluate the DDPM model on upscaled Kanji character images.

Materials

Properties

Simulations

Properties

Simulations

Properties

Simulations

Properties

Simulations