Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Analyzing the Effect of Combined Degradations on Face Recognition

Erdi Sarıtaş, Hazım Kemal Ekenel

TL;DR

This work addresses the robustness of face recognition under real-world degradations by analyzing both single and combined synthetic degradations using an extended exposure-inclusive pipeline applied to LFW. The study employs ArcFace with a ResNet100 backbone and evaluates verification accuracy under normal and cross verification modes, revealing that combined degradations drastically reduce performance compared to single degradations and can lead to near-random accuracy in extreme cases. The findings highlight the necessity of incorporating realistic, multi-factor degradations into robustness assessments and guide future work toward more complex degradation models and ordering. Overall, the paper emphasizes real-world complexity when evaluating face-recognition systems and provides a publicly available pipeline to reproduce and extend the analysis.

Abstract

A face recognition model is typically trained on large datasets of images that may be collected from controlled environments. This results in performance discrepancies when applied to real-world scenarios due to the domain gap between clean and in-the-wild images. Therefore, some researchers have investigated the robustness of these models by analyzing synthetic degradations. Yet, existing studies have mostly focused on single degradation factors, which may not fully capture the complexity of real-world degradations. This work addresses this problem by analyzing the impact of both single and combined degradations using a real-world degradation pipeline extended with under/over-exposure conditions. We use the LFW dataset for our experiments and assess the model's performance based on verification accuracy. Results reveal that single and combined degradations show dissimilar model behavior. The combined effect of degradation significantly lowers performance even if its single effect is negligible. This work emphasizes the importance of accounting for real-world complexity to assess the robustness of face recognition models in real-world settings. The code is publicly available at https://github.com/ThEnded32/AnalyzingCombinedDegradations.

Analyzing the Effect of Combined Degradations on Face Recognition

TL;DR

This work addresses the robustness of face recognition under real-world degradations by analyzing both single and combined synthetic degradations using an extended exposure-inclusive pipeline applied to LFW. The study employs ArcFace with a ResNet100 backbone and evaluates verification accuracy under normal and cross verification modes, revealing that combined degradations drastically reduce performance compared to single degradations and can lead to near-random accuracy in extreme cases. The findings highlight the necessity of incorporating realistic, multi-factor degradations into robustness assessments and guide future work toward more complex degradation models and ordering. Overall, the paper emphasizes real-world complexity when evaluating face-recognition systems and provides a publicly available pipeline to reproduce and extend the analysis.

Abstract

A face recognition model is typically trained on large datasets of images that may be collected from controlled environments. This results in performance discrepancies when applied to real-world scenarios due to the domain gap between clean and in-the-wild images. Therefore, some researchers have investigated the robustness of these models by analyzing synthetic degradations. Yet, existing studies have mostly focused on single degradation factors, which may not fully capture the complexity of real-world degradations. This work addresses this problem by analyzing the impact of both single and combined degradations using a real-world degradation pipeline extended with under/over-exposure conditions. We use the LFW dataset for our experiments and assess the model's performance based on verification accuracy. Results reveal that single and combined degradations show dissimilar model behavior. The combined effect of degradation significantly lowers performance even if its single effect is negligible. This work emphasizes the importance of accounting for real-world complexity to assess the robustness of face recognition models in real-world settings. The code is publicly available at https://github.com/ThEnded32/AnalyzingCombinedDegradations.
Paper Structure (7 sections, 3 equations, 7 figures, 1 table)

This paper contains 7 sections, 3 equations, 7 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. Related Works
  3. METHODOLOGY
  4. EXPERIMENTS
  5. RESULTS
  6. CONCLUSIONS AND FUTURE WORKS
  7. ACKNOWLEDGMENTS

Figures (7)

  • Figure 1: Illustration of degradation pipeline. First, under/over-exposure is applied. Then, a (Gaussian) kernel is used for blurring. Following, (bicubic) downscaling is used and additive (Gaussian) noise is added. Lastly, JPEG artifacts are inserted.
  • Figure 2: The $k_{blur}$ kernels with their parameters on top.
  • Figure 3: Exposure setting results. Examples of single (top) and combined (bottom) degradation results are shown in the below sub-figure.
  • Figure 4: Blurring results. Examples of single (top) and combined (bottom) degradation results are shown in the below sub-figure.
  • Figure 5: Downscaling results. Examples of single (top) and combined (bottom) degradation results are shown in the below sub-figure.
  • ...and 2 more figures