A new preprint on arXiv addresses a practical gap: measured accuracy, speed, and energy of current depth estimators on the Jetson AGX Orin, a common computer for robots and vehicles, are hard to find. The paper, titled "Which Depth Estimator for Jetson AGX Orin? A 2026 Benchmark of Monocular and Stereo Methods," benchmarks nine model families as of October 2026.
The benchmark covers both monocular and stereo depth-estimation methods, according to the title and abstract. The abstract is truncated, so specific results and rankings are not available from the source. The paper's contribution is the benchmark itself, providing comparable measurements on this embedded platform.
Because there is only one source, there are no other findings to compare or contrast. Readers looking for a particular model's performance will need to consult the full preprint for details beyond the abstract.