SP-NET: One Shot Fingerprint Singular-Point Detector

Aditya Nigam; Akash Agrawal; Geetika Arora; Kamlesh Tiwari; Ranjeet Ranjan Jha

REVIEW 4 major objections 5 minor 21 references

SP-NET: One Shot Fingerprint Singular-Point Detector

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read SP-NET detects fingerprint singular points end-to-end and reports higher true detection rates than four prior methods on three databases.

desk verdict A solid architecture paper whose headline SOTA claim is undercut by a sloppy evaluation protocol; worth sending to peer review but needs major revision. read the letter →

arxiv 1908.04842 v1 pith:VMXNWN4F submitted 2019-08-13 cs.CV

Geetika Arora , Ranjeet Ranjan Jha , Akash Agrawal , Kamlesh Tiwari , Aditya Nigam This is my paper · ORCID

classification cs.CV

keywords fingerprintsingularpointdetectiondeeplearningmacro-localizationnetworkmicro-regressionstackedhourglasstruerateFVC2002FPL30K

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

SP-NET is a single end-to-end convolutional network that takes a fingerprint image and returns the pixel coordinates of its singular point, the high-curvature landmark used to normalize fingerprints and drive reliable indexing and matching. The paper's central claim is that this two-part architecture — a macro-localization network that segments a likely region, followed by a micro-regression network that predicts exact coordinates — detects singular points more accurately than existing classical and deep baselines. On FVC2002 DB1_A, FVC2002 DB2_A, and an in-house 30,000-image database FPL30K, the reported true detection rates are 98.75%, 97.5%, and 92.72% at a 20-pixel tolerance, with 96.25% on DB1_A at the stricter 10-pixel tolerance.

What carries the argument

The load-bearing mechanism is the two-stage stack: a Macro-Localization Network, an encoder-decoder with three stacked hourglass bottlenecks that outputs a probability mask, and a Micro-Regression Network, a convolutional block followed by fully connected layers that regresses a single (x,y) coordinate. The hourglass—a multi-scale convolutional module that repeatedly downsamples and upsamples while preserving information through skip connections—lets the network re-assess features across the whole image and across scale, which the paper credits for robustness to noise and blur. The two networks are trained separately and then stacked so a fingerprint image passes through the whole system to produce coordinates in one go.

What would settle it

Re-run the four listed baselines and SP-NET on the same FVC2002 database and identical ground truth under both the 10-pixel and 20-pixel rules; if any baseline's true detection rate at the same tolerance equals or exceeds SP-NET's (98.75% on DB1_A, 96.25% at 10 pixels), the paper's central claim is disproved.

Watch

Extended reading notes

Core claim

The discovery the paper argues for is that singular-point detection can be cast as a regression problem solved in one forward pass, with no hand-designed orientation-field or Poincare-index computation. The Macro-Localization Network uses an encoder-decoder with a three-stacked hourglass bottleneck to produce a 43 × 43-pixel mask around the probable singular point; the Micro-Regression Network then concatenates that mask with the original image and regresses the (x,y) coordinates. Trained with binary cross-entropy for the mask and mean squared error for the coordinates, the combined model reaches the reported true detection rates, which the paper compares favorably against four prior methods in its Table 1.

Load-bearing premise

The load-bearing premise is that Table 1's comparison is fair: it lists SP-NET at a 10-pixel tolerance while not stating the tolerance for the four baselines, and its database caption does not match the text, so if the baselines were scored at the looser 20-pixel rule the 'outperforms all' conclusion would not follow.

Editorial extensions

If this is right

Fingerprint normalization and indexing pipelines can consume singular-point coordinates directly from one forward pass of SP-NET, replacing separate orientation-field and Poincare-index stages.
The architecture can be trained from scratch on a new sensor population: the 92.72% true detection rate on FPL30K, collected from rural subjects across three sensors, indicates the method transfers beyond the clean FVC setting.
At the stricter 10-pixel tolerance the model still reports 96.25% on FVC2002 DB1_A, so its localization error is not merely within the 20-pixel acceptance radius.
Because adding a fourth stacked hourglass gave little improvement, the paper's three-hourglass bottleneck is a deliberate capacity-versus-performance trade-off.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension the paper does not run is an ablation on the mask size: varying the 43 × 43 ground-truth mask while keeping the regressor fixed would show how much of the accuracy is carried by macro-localization versus micro-regression.
The same two-stage mask-plus-regression design could be retrained for other high-curvature landmarks, such as iris or palmprint core points, since nothing in the architecture is fingerprint-specific.
The manual ground-truth labels on FPL30K may set the practical ceiling for reported accuracy; comparing SP-NET's errors against labeler disagreement would separate model error from label noise, a hypothesis the paper leaves implicit.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

The short version: this is a sensible one-shot singular-point detector architecture, but the paper's headline claim — that it beats all state-of-the-art — is not supported by the evaluation as written. The architecture is actually the most interesting part: a U-Net encoder-decoder with three stacked hourglass modules as a bottleneck, followed by a small regression head. That's a legitimate new configuration of known components, and training from scratch without pre-trained weights is a plus in this domain. The macro-localization then micro-regression design is reasonable, and the distance-vs-accuracy curves in Figure 6 give a more honest view than a single TDR number.

The soft spots are in the evaluation protocol, not the architecture. First, Table 1's caption says FVC2002 DB2_A, but Section 3.4 says the 96.25% figure is for DB1_A at 10 pixels. The baselines (Zhou, Xie, Tiwari, Liu) have no stated tolerance, while Eq. 3 defines TDR at 20 pixels. That is a mess. If the baselines were evaluated at 20 pixels, SP-NET's 10-pixel number is a stricter result, which could actually strengthen the comparison, but the mismatch needs to be fixed and the protocol stated precisely. Second, and more concerning, the 80-20 split is not stated to be subject-disjoint. FVC2002 has multiple impressions of the same finger; a random image split lets the same finger appear in both training and test, which can inflate TDR substantially. This is the kind of leak that can turn a 2% gain into a phantom. Third, the network outputs a single (x,y) coordinate, but fingerprints routinely have multiple singularities (core and deltas). The paper never says whether it only targets the core or defines a primary singular point, so the metric's meaning is unclear. These are fixable, but they are real.

The paper also doesn't release code or the in-house FPL30K ground truth, which limits reproducibility. That's a common downside for this kind of applied work, but worth noting.

Who should read this: people working on fingerprint indexing or landmark localization in structured images. The architecture idea is worth taking seriously, and the evaluation pitfalls are a useful cautionary example.

My recommendation: send it to peer review. It deserves referee time, but the authors need to do a major revision — clean up the table, specify the tolerance for every method, redo the split subject-disjoint, and clarify how multiple singularities are handled. As written, the central claim of superiority should be treated as unverified.

Referee Report

4 major / 5 minor

Summary. The paper proposes SP-NET, a deep CNN for fingerprint singular-point detection. The architecture combines a Macro-Localization Network (an encoder-decoder with three stacked hourglass modules) and a Micro-Regression Network that outputs singular-point coordinates. The model is trained separately on the segmentation and regression stages and then stacked into a single forward-pass detector. The authors report true detection rates (TDR) of 98.75%, 97.5%, and 92.72% on FVC2002 DB1_A, FVC2002 DB2_A, and an in-house FPL30K dataset (30,000 images), respectively, and claim that this outperforms all state-of-the-art techniques. The paper includes a distance-vs-accuracy analysis and a comparison with four prior methods on one database.

Significance. If the reported results hold under a properly controlled protocol, the paper would represent a useful contribution: an end-to-end CNN for singular-point localization that avoids hand-crafted orientation-field features, achieves high TDR on standard benchmarks, and introduces a large manually annotated in-house dataset. The two-stage macro-localization/micro-regression design with stacked hourglass modules is technically reasonable, and the distance-vs-accuracy plots provide more information than a single TDR value. The central claim of state-of-the-art superiority is, however, not currently supported because the comparative evaluation in Table 1 is internally inconsistent and does not establish a common protocol for the baselines. The paper also does not release code or trained models, which limits independent verification.

major comments (4)

[Section 3.4, Table 1] The dataset label is inconsistent between text and table. Section 3.4 states that the proposed model shows a detection rate of 96.25% on FVC2002 DB1_A at a 10-pixel tolerance, but the caption of Table 1 says the comparison is on FVC2002 DB2_A. Table 2 reports a 20-pixel TDR of 98.75% for DB1_A and 97.50% for DB2_A, which indicates that the 96.25% figure in Table 1 corresponds to DB1_A, not DB2_A. This mismatch must be corrected before the comparison can be interpreted.
[Table 1, Equation 3] The comparison with baselines in Table 1 is not protocol-equivalent. Equation 3 defines a true detection as a Euclidean distance of at most 20 pixels, and Table 2 uses that threshold. The proposed SP-NET row in Table 1, however, is explicitly reported 'at less than 10 pixels', while the rows for Zhou, Xie, Tiwari, and Liu do not state the tolerance, ground-truth definition, or test split used. If the baselines were evaluated under the more lenient 20-pixel rule, or on a different ground truth, the claimed superiority is not established by the table as presented. The authors should either re-run the baselines under the same protocol or clearly report each method's original tolerance and explain the comparability.
[Section 3.4, Table 2] The reported TDR values come from a single 80%-20% train-test split with no cross-validation, confidence intervals, or error bars. On FVC2002 DB1_A and DB2_A, the test sets contain only 160 images each, so a change of one or two predictions changes the TDR by roughly 0.6-1.2 percentage points. The difference between the proposed method's 10-pixel TDR (96.25%) and the next-best baseline (96.03%) is within this sampling variability. The authors should report results over multiple splits or provide bootstrap confidence intervals to support the claimed margin of improvement.
[Section 3.1, Abstract] The abstract claims that SP-NET is 'better than any other state-of-the-art technique' on all three databases, but state-of-the-art comparisons are provided only for one database (and even that comparison is protocol-inconsistent). FPL30K is an in-house dataset with no public release, and the paper gives no baseline results on it. The broad superiority claim is therefore not supported by the evidence presented. The authors should either add controlled comparisons on all databases or substantially soften the claim.

minor comments (5)

[Title, Abstract] The term 'one-shot' is potentially misleading. The model performs detection in a single forward pass, but it is not one-shot learning in the standard machine-learning sense. Please clarify this terminology.
[Section 3.1] The ground truth for all databases was 'generated manually', but no details are given about the annotation protocol, the number of annotators, inter-annotator agreement, or quality control. This is important for the credibility of the TDR numbers, especially for the 30,000-image FPL30K dataset.
[Section 3.2] The 20-pixel tolerance in Equation 3 is introduced without justification. The distance-vs-accuracy curves in Figure 6 partially address this, but the choice of 20 pixels as the headline criterion should be discussed or referenced.
[Section 2.2] The abstract and introduction describe the model as 'end-to-end', but Section 2.2 says the two subnetworks are trained individually and then stacked. It is unclear whether the stacked network is subsequently fine-tuned as a whole. Please clarify the training procedure.
[Section 1.2] The contribution list claims robustness to 'scale difference, blurriness etc.' due to repeated scale-up/down operations, but no experiments on scale or blur variations are presented. This claim is not empirically supported.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: SP-NET reports trained-model measurements against manually labeled ground truth; reported TDRs are empirical outputs, not fitted inputs.

full rationale

The paper claims an end-to-end CNN that localizes fingerprint singular points and reports TDR on FVC2002 DB1_A, FVC2002 DB2_A, and FPL30K. The derivation chain is architectural and empirical: the macro-localization network is trained with binary cross-entropy on manually generated masks, the micro-regression network is trained with mean squared error on manually marked coordinates, and TDR is defined in Eq. 3 as a Euclidean distance from predicted to ground-truth coordinates. No parameter of the model is fit to the TDR numbers, and no reported quantity is defined in terms of another reported quantity. The ground truth is generated manually (Section 3.1), so the evaluation is not self-referential. Self-citations appear (e.g., [17] in Table 1 and [4] in the introduction), but they are baseline numbers and background references, not load-bearing justifications for the method's correctness; the cited numbers are externally published results. Table 1's mismatched database label and the unstated tolerance for baseline methods are legitimate soundness and protocol concerns, but they are not circularity: they concern comparability, not the reduction of a prediction to its inputs. Therefore no circular step can be quoted, and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim is an empirical performance measurement. The free parameters listed (tolerance, mask size, number of hourglasses) are choices that shape the reported numbers. The axioms are the unstated premises about ground truth quality and data splitting that the evaluation relies on.

free parameters (3)

Tolerance threshold for TDR = 20 pixels (Eq. 3); 10 pixels in Table 1
The true detection rate definition in Eq. 3 uses a 20-pixel Euclidean distance cutoff. This threshold is chosen by the authors and directly determines the reported TDR values. Table 1 reports SP-Net at 10 pixels, creating an inconsistent comparison.
Ground truth mask size = 43x43 pixels
The mask for the singular point region is a 43x43 square around the manually marked ground truth (Section 3.1). This choice affects the MLN training target.
Number of stacked hourglass modules = 3
Section 2.1 states that three hourglass networks were chosen after testing one to four; 'there was not much improvement when four' were used. This is a hand-tuned architecture choice.

assumptions (3)

domain assumption Manually annotated ground truth coordinates are accurate and consistent
The paper states the ground truth for all databases was generated manually (Section 3.1). No inter-annotator agreement or validation against an independent method is reported. The entire TDR measurement depends on these annotations.
domain assumption The 80-20 train/test split is representative and there is no subject overlap leakage
Section 2.2 states each dataset is divided into 80%-20% for training and testing. For FVC2002, each subject has 8 impressions; if the split is per-image rather than per-subject, impressions from the same finger could appear in both train and test, inflating accuracy. The paper does not specify that the split is subject-disjoint.
standard math Standard deep learning assumptions (i.i.d. data, convergence of Adam optimizer)
Backpropagation and optimization are assumed to work as usual; no proof given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SP-NET: One Shot Fingerprint Singular-Point Detector." pith.science (2026). https://pith.science/paper/VMXNWN4F

@misc{pith2026190804842,
  author       = {Pith},
  title        = {Pith review of: SP-NET: One Shot Fingerprint Singular-Point Detector},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VMXNWN4F}},
  note         = {Machine review of arXiv:1908.04842}
}

read the original abstract

Singular points of a fingerprint image are special locations having high curvature properties. They can play a pivotal role in fingerprint normalization and reliable feature extraction. Accurate and efficient extraction of a singular point plays a major role in successful fingerprint recognition and indexing. In this paper, a novel deep learning based architecture is proposed for one shot (end-to-end) singular point detection from an input fingerprint image. The model consists of a Macro-Localization Network and a Micro-Regression Network along with three stacked hourglass as a bottleneck. The proposed model has been tested on three databases viz. FVC2002 DB1_A, FVC2002 DB2_A and FPL30K and has been found to achieve true detection rate of 98.75%, 97.5% and 92.72% respectively, which is better than any other state-of-the-art technique.

Figures

Figures reproduced from arXiv: 1908.04842 by the authors.

**Figure 2.** Block Diagram of Proposed SP-Net (dotted lines shows logical parts) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Architecture of Proposed Macro Localization Network [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Hourglass Network Used in Proposed Model (Dotted lines represents skip connec [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Architecture of Proposed Micro Regression Network [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: Distance vs Accuracy plots on FPL30K, FVC2002 DB1_A and FVC2002 DB2_A [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 20 canonical work pages

[1]

http://bias.csr.unibo.it/fvc2002/

The FVC2002 database:. http://bias.csr.unibo.it/fvc2002/

work page
[2]

Fingerprint singular point detection algorithm by poincaré index

Jin Bo, Tang Hua Ping, and Xu Ming Lan. Fingerprint singular point detection algorithm by poincaré index. WSEAS Transactions on Systems, 7(12):1453–1462, 2008

work page 2008
[3]

A fast and accurate method for detecting ﬁngerprint reference point

Xifeng Guo, En Zhu, and Jianping Yin. A fast and accurate method for detecting ﬁngerprint reference point. Neural Computing and Applications, 29(1):21–31, 2018

work page 2018
[4]

Fingerprint indexing schemes - a survey

Phalguni Gupta, Kamlesh Tiwari, and Geetika Arora. Fingerprint indexing schemes - a survey. Neurocomputing, 335:352–365, 2019. 10 G ARORA, RR JHA ET AL.: SP-NET: FINGERPRINT SINGULAR-POINT DETECTOR

work page 2019
[5]

A robust singular point detection algorithm

Puneet Gupta and Phalguni Gupta. A robust singular point detection algorithm. Applied Soft Computing, 29:411–423, 2015

work page 2015
[6]

Fingerprint singular point de- tection based on modiﬁed poincare index method

Gabriel Babatunde Iwasokun and Oluwole Charles Akinyokun. Fingerprint singular point de- tection based on modiﬁed poincare index method. International Journal of Signal Processing, Image Processing and Pattern Recognition, 7(5):259–272, 2014

work page 2014
[7]

Handbook of biometrics

Anil K Jain, Patrick Flynn, and Arun A Ross. Handbook of biometrics . Springer Science & Business Media, 2007

work page 2007
[8]

Algorithm of extracting core point in ﬁngerprint.JOURNAL- HEBEI UNIVERSITY OF TECHNOLOGY, 36(1):46, 2007

L Liang, H Zhao, P He, and H Tian. Algorithm of extracting core point in ﬁngerprint.JOURNAL- HEBEI UNIVERSITY OF TECHNOLOGY, 36(1):46, 2007

work page 2007

Show all 21 references

[9]

A method for singular points detection based on faster-rcnn

Yonghong Liu, Baicun Zhou, Congying Han, Tiande Guo, and Jin Qin. A method for singular points detection based on faster-rcnn. Applied Sciences, 8(10):1853, 2018

2018
[10]

Handbook of ﬁngerprint recognition

Dario Maio and Anil K Jain. Handbook of ﬁngerprint recognition. springer, 2009

2009
[11]

Fingerprint core location algorithm based on sliding window

Xiangshen Min, Xuefeng Zhang, and Fang Ren. Fingerprint core location algorithm based on sliding window. Wuhan University Journal of Natural Sciences, 23(3):195–200, 2018

2018
[12]

Stacked hourglass networks for human pose esti- mation

Alejandro Newell, Kaiyu Yang, and Jia Deng. Stacked hourglass networks for human pose esti- mation. In European Conference on Computer Vision, pages 483–499. Springer, 2016

2016
[13]

Ipsegnet: Deep convolutional neural network based segmentation framework for iris and pupil

Shreyas Malakarjun Patil, Ranjeet Ranjan Jha, and Aditya Nigam. Ipsegnet: Deep convolutional neural network based segmentation framework for iris and pupil. In International Conference on Signal-Image Technology & Internet-Based Systems (SITIS), pages 184–191. IEEE, 2017

2017
[14]

Faster r-cnn: Towards real-time object detection with region proposal networks

Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems, pages 91–99, 2015

2015
[15]

U-Net: Convolutional networks for biomedical image segmentation

Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-Net: Convolutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention, pages 234–241. Springer, 2015

2015
[16]

Indexing ﬁngerprint database with minutiae based coaxial gaussian track code and quantized lookup table

Kamlesh Tiwari and Phalguni Gupta. Indexing ﬁngerprint database with minutiae based coaxial gaussian track code and quantized lookup table. In International Conference on Image Process- ing, pages 4773–4777. IEEE, 2015

2015
[17]

Meandering energy potential to locate singular point of ﬁngerprint

Kamlesh Tiwari and Phalguni Gupta. Meandering energy potential to locate singular point of ﬁngerprint. In International Conference on Biometrics (ICB), pages 1–6. IEEE, 2016

2016
[18]

Fingerprint reference point determination based on a novel ridgeline feature

SJ Xie, HM Yoo, DS Park, and S Yoon. Fingerprint reference point determination based on a novel ridgeline feature. In Intl. Conference on Image Processing, pages 3073–3076. IEEE, 2010

2010
[19]

Fingerprint reference point identiﬁcation based on chain encoded discrete curvature and bending energy

Geevar C Zacharias, Madhu S Nair, and P Sojan Lal. Fingerprint reference point identiﬁcation based on chain encoded discrete curvature and bending energy. Pattern Analysis and Applica- tions, 20(1):253–267, 2017

2017
[20]

A novel algorithm for detecting singular points from ﬁngerprint images

Jie Zhou, Fanglin Chen, and Jinwei Gu. A novel algorithm for detecting singular points from ﬁngerprint images. IEEE Transactions on Pattern Analysis and Machine Intelligence , 31(7): 1239–1250, 2009

2009
[21]

Walking to singular points of ﬁngerprints

En Zhu, Xifeng Guo, and Jianping Yin. Walking to singular points of ﬁngerprints. Pattern Recognition, 56:116–128, 2016

2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

http://bias.csr.unibo.it/fvc2002/

The FVC2002 database:. http://bias.csr.unibo.it/fvc2002/

work page

[2] [2]

Fingerprint singular point detection algorithm by poincaré index

Jin Bo, Tang Hua Ping, and Xu Ming Lan. Fingerprint singular point detection algorithm by poincaré index. WSEAS Transactions on Systems, 7(12):1453–1462, 2008

work page 2008

[3] [3]

A fast and accurate method for detecting ﬁngerprint reference point

Xifeng Guo, En Zhu, and Jianping Yin. A fast and accurate method for detecting ﬁngerprint reference point. Neural Computing and Applications, 29(1):21–31, 2018

work page 2018

[4] [4]

Fingerprint indexing schemes - a survey

Phalguni Gupta, Kamlesh Tiwari, and Geetika Arora. Fingerprint indexing schemes - a survey. Neurocomputing, 335:352–365, 2019. 10 G ARORA, RR JHA ET AL.: SP-NET: FINGERPRINT SINGULAR-POINT DETECTOR

work page 2019

[5] [5]

A robust singular point detection algorithm

Puneet Gupta and Phalguni Gupta. A robust singular point detection algorithm. Applied Soft Computing, 29:411–423, 2015

work page 2015

[6] [6]

Fingerprint singular point de- tection based on modiﬁed poincare index method

Gabriel Babatunde Iwasokun and Oluwole Charles Akinyokun. Fingerprint singular point de- tection based on modiﬁed poincare index method. International Journal of Signal Processing, Image Processing and Pattern Recognition, 7(5):259–272, 2014

work page 2014

[7] [7]

Handbook of biometrics

Anil K Jain, Patrick Flynn, and Arun A Ross. Handbook of biometrics . Springer Science & Business Media, 2007

work page 2007

[8] [8]

Algorithm of extracting core point in ﬁngerprint.JOURNAL- HEBEI UNIVERSITY OF TECHNOLOGY, 36(1):46, 2007

L Liang, H Zhao, P He, and H Tian. Algorithm of extracting core point in ﬁngerprint.JOURNAL- HEBEI UNIVERSITY OF TECHNOLOGY, 36(1):46, 2007

work page 2007