Mixing Synthetic and Real Images Improves Artificial Intelligence-Based Detection of the Pupil, Iris, and Sclera: A Cross-Domain Validation Study
Krishna Keshav, Deepsekhar Das, Sumit Grover, Sahil Agrawal, Avinav Bharti · 2026
Read Mixing Synthetic and Real Images Improves Artificial Intelligence-Based Detection of the Pupil, Iris, and Sclera: A Cross-Domain Validation Study by Krishna Keshav, Deepsekhar Das, Sumit Grover, Sahil Agrawal, Avinav Bharti (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13283298.txt.
Description
Krishna Keshav and co-authors tested in 2026 whether mixing made-up and real eye photos improves computer training for spotting eye parts. Models must outline the pupil, iris, and white of the eye across image types to be useful in clinics. Four models were trained: one on generated images only, one on real images only, and two on half-and-half mixes of different sizes.
Single-source models broke down on the other image type, in one case missing a pupil entirely. The mixed model of 100 images reached about 89 percent pupil accuracy with no failures and far steadier results. Doubling the mixed set added nothing. The authors conclude that mixing sources, not adding volume, makes eye software dependable.
Catalog record
| CREATOR | Krishna Keshav, Deepsekhar Das, Sumit Grover, Sahil Agrawal, Avinav Bharti |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |