Self-supervised cross frame-rate event-frame depth learning without ground truth
Arun Kumar Arigela, D. David Neels Ponkumar, Martin Margala, K. Jayasri, A. Leela Sravanthi, Prasun Chakrabarti · 2026
Read Self-supervised cross frame-rate event-frame depth learning without ground truth by Arun Kumar Arigela, D. David Neels Ponkumar, Martin Margala, K. Jayasri, A. Leela Sravanthi, Prasun Chakrabarti (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/PMC13578208.txt.
Description
Arun Kumar Arigela and colleagues present a 2026 method for judging depth from event cameras, sensors that log brightness changes with very fine timing. Such cameras suit fast, hard scenes, yet depth methods often need hand-checked depth maps for training. The team set out to learn depth with no ground truth at all.
Their system pairs event streams with normal frames in one transformer model, a type of network that weighs links across a whole scene. Attention across mismatched frame rates ties the two feeds together over time. Training leans on how well rebuilt images match the real ones, plus smoothness and timing checks. On a standard test set the method beat earlier deep learning systems on the main error scores.
Catalog record
| CREATOR | Arun Kumar Arigela, D. David Neels Ponkumar, Martin Margala, K. Jayasri, A. Leela Sravanthi, Prasun Chakrabarti |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 124248 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |