English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The latest advances in deep learning have facilitated the development ofhighly accurate monocular depth estimation models. However, when training amonocular depth estimation network, practitioners and researchers have observednot a number (NaN) loss, which disrupts gradient descent optimization. Althoughseveral practitioners have reported the stochastic and mysterious occurrence ofNaN loss that bothers training, its root cause is not discussed in theliterature. This study conducted an in-depth analysis of NaN loss duringtraining a monocular depth estimation network and identified three types ofvulnerabilities that cause NaN loss: 1) the use of square root loss, whichleads to an unstable gradient; 2) the log-sigmoid function, which exhibitsnumerical stability issues; and 3) certain variance implementations, whichyield incorrect computations. Furthermore, for each vulnerability, theoccurrence of NaN loss was demonstrated and practical guidelines to prevent NaNloss were presented. Experiments showed that both optimization stability andperformance on monocular depth estimation could be improved by following ourguidelines.

Analysis of NaN Divergence in Training Monocular Depth Estimation Model