# Deep Learning-Based Feature Extraction in Iris Recognition: Use Existing Models, Fine-tune or Train From Scratch?

arXiv: https://arxiv.org/abs/2002.08916 · HTML: https://mlsanity.com/paper/2002.08916/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 20 Feb 2020.

## Built on by (most cited first, 9 of 9)

- [Human-Aided Saliency Maps Improve Generalization of Deep Learning](https://mlsanity.com/paper/2105.03492/) (2021)
- [Iris Liveness Detection using a Cascade of Dedicated Deep Learning Networks](https://mlsanity.com/paper/2105.14009/) (2021)
- [Interpretable Deep Learning-Based Forensic Iris Segmentation and Recognition](https://mlsanity.com/paper/2112.00849/) (2021)
- [Deep Learning for Iris Recognition: A Review](https://mlsanity.com/paper/2303.08514/) (2023)
- [Iris Style Transfer: Enhancing Iris Recognition with Style Features and Privacy Preservation through Neural Style Transfer](https://mlsanity.com/paper/2503.04707/) (2025)
- [AutoSIGHT: Automatic Eye Tracking-based System for Immediate Grading of Human experTise](https://mlsanity.com/paper/2508.01015/) (2025)
- [Privacy Attacks Against Biometric Models with Fewer Samples: Incorporating the Output of Multiple Models](https://mlsanity.com/paper/2209.11020/) (2022)
- [Adaptive Deep Iris Feature Extractor at Arbitrary Resolutions](https://mlsanity.com/paper/2407.08341/) (2024)
- [On the Feasibility of Creating Iris Periocular Morphed Images](https://mlsanity.com/paper/2408.13496/) (2024)

## Abstract (arXiv)

Modern deep learning techniques can be employed to generate effective feature extractors for the task of iris recognition. The question arises: should we train such structures from scratch on a relatively large iris image dataset, or it is better to fine-tune the existing models to adapt them to a new domain? In this work we explore five different sets of weights for the popular ResNet-50 architecture to find out whether iris-specific feature extractors perform better than models trained for non-iris tasks. Features are extracted from each convolutional layer and the classification accuracy achieved by a Support Vector Machine is measured on a dataset that is disjoint from the samples used in training of the ResNet-50 model. We show that the optimal training strategy is to fine-tune an off-the-shelf set of weights to the iris recognition domain. This approach results in greater accuracy than both off-the-shelf weights and a model trained from scratch. The winning, fine-tuned approach also shows an increase in performance when compared to previous work, in which only off-the-shelf (not fine-tuned) models were used in iris feature extraction. We make the best-performing ResNet-50 model, fine-tuned with more than 360,000 iris images, publicly available along with this paper.