On the impact of deep neural network calibration on adaptive edge offloading for image classification

Roberto G. Pacheco, Rodrigo S. Couto, Osvaldo Simeone

Research output: Contribution to journalArticlepeer-review

2 Scopus citations


Edge devices can offload deep neural network (DNN) inference to the cloud to overcome energy or processing constraints. Nevertheless, offloading adds communication delay, which increases the overall inference time. An alternative is to use adaptive offloading based on early-exit DNNs. Early-exit DNNs have branches inserted at the output of given intermediate layers. These side branches provide confidence estimates. If the confidence level of the decision produced is sufficient, the inference is made by the side branch. Otherwise, the edge offloads the inference decision to the cloud, which implements the remaining DNN layers. Thus, the offloading decision depends on reliable confidence levels provided by the side branches at the device. This article provides an extensive calibration study on different datasets and early-exit DNNs for the image classification task. Our study shows that early-exit DNNs are often miscalibrated, overestimating their prediction confidence and making unreliable offloading decisions. To evaluate the impact of calibration on accuracy and latency, we introduce two novel application-level metrics and evaluate well-known DNN models in a realistic edge computing scenario. The results demonstrated that calibrating early-exit DNNs improves the probabilities of meeting accuracy and latency requirements.

Original languageEnglish (US)
Article number103679
JournalJournal of Network and Computer Applications
StatePublished - Aug 2023

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Computer Science Applications
  • Computer Networks and Communications


  • Deep neural network calibration
  • Early-exit deep neural networks
  • Edge computing
  • Edge offloading


Dive into the research topics of 'On the impact of deep neural network calibration on adaptive edge offloading for image classification'. Together they form a unique fingerprint.

Cite this