Dynamic convolutional neural networks for altitude aware UAV object detection
Michalis Piponidis, Theocharis Theocharides · 2026
Read Dynamic convolutional neural networks for altitude aware UAV object detection by Michalis Piponidis, Theocharis Theocharides (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/PMC13562726.txt.
Description
Michalis Piponidis and Theocharis Theocharides argue that a drone's object detector should reshape itself to match the height it is flying at. Objects shrink and spread as altitude changes, yet one fixed network wastes scarce onboard power by treating every view the same. Their 2026 work trains several lean network settings, each tuned for a band of altitudes, varying image size, network width, and filter size. A threshold rule then swaps settings in flight as the drone climbs or descends.
Two datasets rich in altitude variety confirmed that the best settings differ sharply between low and high flight. With well-chosen thresholds, the switching system lost almost no accuracy against a single heavyweight network, while saving large amounts of computation and memory. Poorly placed thresholds, the analysis shows, erode that bargain. The authors pitch altitude-aware switching for surveillance, rescue, and environmental patrol drones that must think fast on small batteries.
Catalog record
| CREATOR | Michalis Piponidis, Theocharis Theocharides |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65934 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |