Abstract
Sphere Encoder is an autoencoder that generates images by decoding random points from a high-dimensional latent sphere. We identify two limitations of the original formulation that reduce its generation quality. First, random points concentrate near the equator relative to the pole on an encoded latent, but the training rotation never reaches this region, leaving a gap that limits one-step generation. Second, training for generation with pixel-wise reconstruction loss encourages the decoder to average over plausible images, producing blurry images that lack high-frequency details. We present Sphere Encoder 2 to address both limitations, substantially improving image generation quality while maintaining the speed and simplicity of a autoencoder. Models are released at https://github.com/kaiyuyue/sphere2.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Yue, K., McLeish, S., Rawal, R., Bartoldson, B., Jia, M., & Goldstein, T. (2026). Sphere Encoder 2. https://omanscience.com/en/articles/sphere-encoder-2
MLA 9
Yue, Kaiyu, et al. "Sphere Encoder 2." https://omanscience.com/en/articles/sphere-encoder-2.
Chicago (author–date)
Yue, Kaiyu, Sean McLeish, Ruchit Rawal, Brian Bartoldson, Menglin Jia, and Tom Goldstein. 2026. "Sphere Encoder 2." https://omanscience.com/en/articles/sphere-encoder-2.
Harvard
Yue, K., McLeish, S., Rawal, R., Bartoldson, B., Jia, M. and Goldstein, T. (2026) 'Sphere Encoder 2', Available at: https://omanscience.com/en/articles/sphere-encoder-2.
Vancouver
Yue K, McLeish S, Rawal R, Bartoldson B, Jia M, Goldstein T. Sphere Encoder 2. https://omanscience.com/en/articles/sphere-encoder-2
IEEE
K. Yue, S. McLeish, R. Rawal, B. Bartoldson, M. Jia, and T. Goldstein, "Sphere Encoder 2," https://omanscience.com/en/articles/sphere-encoder-2.