Pith. sign in

REVIEW 4 major objections 4 minor 54 references

Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks

T0 review · 4 major / 4 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 FGVis produces fine-grained, class-discriminative explanations directly in image space by clipping gradients that would let the explanation activate neurons beyond the range seen in the original image.

desk verdict FGVis delivers a genuinely new gradient-clipping defense and fine-grained image-space masks, but the defense's guarantee is overstated — Eq. 4 permits deletion-based adversarial evidence that Table 1 never tests. read the letter →

arxiv 1908.02686 v1 pith:GAZU2ZSP submitted 2019-08-07 cs.CV cs.LG

Jörg Wagner , Jan Mathias Köhler , Tobias Gindele , Leon Hetzel , Jakob Thaddäus Wiedemer , Sven Behnke This is my paper · ORCID
classification cs.CVcs.LG
keywords visualexplanationsconvolutionalneuralnetworksadversarialevidencegradientclippingfine-grainedsaliencyperturbation-basedweakly-supervisedlesiondetectionmodelfaithfulness
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

This paper proposes FGVis, a post-hoc method that explains a CNN's prediction by optimizing a sparse mask that removes pixels from the image, producing an explanation that is itself a valid model input. The paper's central technical claim is that adversarial evidence -- spurious patterns introduced by the optimization that make the explanation look confident for a class that is visually absent -- can be blocked by clipping gradients in the backward pass so that no neuron's activation in the explanation may exceed the activation range it reaches on the original image. Because the defense is parameter-free and imposes no smoothness or resolution constraint on the mask, the resulting explanations preserve edges and colors and remain fine-grained down to individual pixels. The paper reports that FGVis outperforms Grad-CAM, LIME, RISE, and sliding-window occlusion on the deletion faithfulness metric on ImageNet, and that it matches or beats fully supervised lesion-detection methods on the DiaretDB1 diabetic retinopathy dataset even though it is only weakly supervised. If the method works as described, practitioners would gain testable pixel-level evidence of what a network sees, including small structures such as microaneurysms, and a way to expose biases such as reliance on color.

What carries the argument

The central object is the backward-pass gradient-clipping rule of Eq. 6, which is equivalent to inserting a clipping layer after each nonlinearity whose forward pass is the identity. For every neuron $i$ in layer $l$, the layer stores bounds $b_u = \max(0, h_i^l(x))$ and $b_l = \min(0, h_i^l(x))$ computed on the original image $x$, and in the backward pass the incoming gradient is multiplied by the indicator that the explanation's activation $h_i^l(e_{c_T})$ lies inside $[b_l, b_u]$. This enforces the constraint of Eq. 4 -- the explanation may activate a neuron no more strongly than the original image did -- while leaving the forward model untouched, adding no hyperparameters, and imposing no smoothness or resolution limit on the mask. The mask optimization of Eq. 2 and Eq. 3 with a zero reference image then yields sparse, pixel-level explanations that are valid model inputs.

What would settle it

Run the defended optimization (generation game with $\lambda=0$) on a large held-out set of images, targeting a class that is visually absent, and count how often the resulting explanation is classified as that class with softmax score above a fixed threshold such as 10%; the paper reports near-zero success on 1000 ImageNet images, so any material fraction of successes would refute the claim that the defense blocks adversarial evidence. A second test would construct an input and target class where legitimate evidence requires a neuron to fire more strongly than it does in the original image, and check whether the per-location clipping suppresses that evidence; if it does, the constraint is too strong for fine-grained explanations.

Watch

Extended reading notes

Core claim

The paper's central claim is that optimization-based visual explanations can be made both fine-grained and free of adversarial evidence without any human-tuned defense parameters. The mechanism is a backward-pass gradient clip: after every nonlinear layer, the explanation's activation for each neuron is silently constrained to the range that the same neuron attains on the original input, which the paper justifies by the pruning convention that zero activations carry the least information. In effect the feature set the explanation may use is a subset of the features already present in the image, so the optimizer cannot sculpt brand-new evidence. The paper validates the defense by trying to generate explanations for the least-likely ImageNet class: without the defense an adversarial explanation is produced for 100% of images, and with the defense for roughly 0% across VGG16, AlexNet, ResNet50, and GoogleNet. Built on this defense, FGVis computes deletion and preservation explanations directly in image space and, to the authors' knowledge, is the first method able to produce fine-grained explanations directly in the image space.

Load-bearing premise

The defense assumes that a neuron whose activation is zero carries the least information, so that preventing any neuron from firing more strongly in the explanation than it does in the original image is enough to stop the optimizer from inventing new evidence; if that premise fails, the clipping would either admit some adversarial artifacts or filter out legitimate fine-grained evidence that needs a neuron to exceed its original activation.

Editorial extensions

If this is right

  • Explanations from FGVis are valid model inputs, so faithfulness metrics such as deletion curves can be computed directly on the model without proxy measures or preprocessing that could distort the result.
  • The parameter-free defense removes the need to tune smoothness or resolution regularizers per model, making fine-grained explanations practical across different CNN architectures.
  • In medical imaging, weakly supervised localization of small lesions -- such as red small dots in diabetic retinopathy -- reaches sensitivities comparable to supervised detectors designed for a single lesion type.
  • The method can expose training-data biases: preservation explanations of school buses consistently feature yellow, and swapping color channels collapses the model's accuracy for color-dependent classes.
  • The defense gives a simple verification check: if the model is confident for a class that is visually absent, the explanation should be nearly empty, so users can test an explanation's credibility directly.

Reading between the lines

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

  • Because the constraint is applied per neuron at a fixed spatial location, FGVis implicitly treats features as location-bound; a testable extension would allow location-shifted evidence using the network's own translational invariances and check whether that can be added without reintroducing adversarial artifacts.
  • The zero-activation premise is the most promising attack surface: an adversary could try to encode a class through a combination of already-active neurons, none of which individually exceeds its original bound, which would bypass the defense without violating Eq. 4.
  • The color-bias analysis suggests a systematic audit recipe: compute preservation explanations for a class, swap color channels, then measure both the accuracy drop and the shift in explanation color; applied broadly, that recipe could reveal dataset biases that accuracy alone hides.
  • On DiaretDB1 the paper compares FGVis at a 25% overlap threshold against a published CAM baseline reported at 50% overlap; a head-to-head with identical thresholds would separate a genuine fine-grained-localization advantage from a metric artifact.
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.

Referee Report

4 major / 4 minor

Summary. The paper proposes FGVis, a post-hoc optimization-based visual explanation method that produces fine-grained, image-space explanations by optimizing a mask-based perturbation of the input. Its main technical contribution is an adversarial-defense technique that clips gradients during backpropagation whenever a hidden activation would exceed the range of activations produced by the original image (Eqs. 4–6). The defense is designed to prevent the optimization from hallucinating adversarial evidence while avoiding human-tuned hyperparameters. The authors evaluate the defense on an adversarial-generation task (Table 1), compare faithfulness against Grad-CAM, LIME, RISE, and sliding-window occlusion using the deletion metric (Table 2), and apply FGVis to weakly supervised lesion detection in diabetic retinopathy fundus images (Table 3). The paper also presents qualitative results on class discriminability, fine-grained detail, and training-data color bias.

Significance. If the central claims held as stated, the work would be valuable: it offers a parameter-free mechanism for constraining explanation optimization to use only evidence present in the original image, which would be a practically useful alternative to low-resolution masking or stochastic defenses. The derivation of the backward-pass clipping rule (Eq. 6) is clean and readily implementable, and the adversarial-generation experiment in Table 1 is a useful stress test showing that the defense blocks additive hallucination. The medical localization study goes beyond toy benchmarks. However, the paper's core guarantee that the defense prevents adversarial evidence is stronger than what Eq. 4 actually enforces, and several quantitative comparisons are not evaluated on equal terms. These issues bear directly on the paper's main contribution, so the manuscript needs substantive revision and additional experiments rather than cosmetic changes.

major comments (4)
  1. [§3.2, Eqs. (4)–(6)] The claim that the activation-bounding constraint prevents adversarial evidence is too strong. For a ReLU network, Eq. 4 allows any activation h_i^l(e) to decrease to zero even when h_i^l(x) > 0. Because the logit for a target class c_A is linear in the top-layer activations, logit_{c_A}(e) = Σ_i W_{c_A,i} h_i^L(e) + b_{c_A}, deleting a feature that is active in x and has a negative weight for c_A increases logit_{c_A}(e). The feasible set of Eq. 4 therefore contains explanations for a visually absent class c_A that are obtained by deleting existing evidence, not by generating new evidence. Table 1 and Table A1 validate the defense only with the generation game starting from a zero image, so the defense's guarantee against deletion-based adversarial evidence is not established by the reported experiments. Please either extend the validation to the deletion/repression games or explicitly restrict the claim to additive hallucination.
  2. [§5.1, Table 2] The faithfulness evaluation is partly circular. The deletion metric measures how quickly the target-class probability drops when pixels are removed in order of an importance map, and that importance map is exactly the mask produced by the deletion game (Eq. 3), which optimizes a sparsity-weighted objective to drop the target-class probability. A method that directly optimizes this objective is expected to score well on the deletion metric relative to methods that are not trained against it. The paper should report an independent faithfulness measure (e.g., an insertion metric or a human-grounded evaluation) or at least discuss this optimization-to-metric alignment as a limitation before claiming that FGVis 'outperforms' the reference methods by a large margin.
  3. [§5.2, Table 3] The comparison with Gondal et al. is not threshold-matched: FGVis is evaluated with a 25% overlap criterion, whereas the reference weakly supervised method uses a 50% overlap criterion. At the 50% overlap that is also reported in Table 3, FGVis is below Gondal et al. for H and HE and comparable for SE, so the statement that FGVis 'performs comparable or outperforms fully supervised approaches' at image-level lesion detection is not supported when the same evaluation criterion is used. The headline comparison should be made at the same threshold, or the claim should be qualified accordingly.
  4. [§3.2, Eq. (6)] The forward pass is not clamped, so Eq. 4 is not actually enforced at any iteration or at the final explanation. Equation 6 only zeros gradients whose update would violate the bounds; an SGD step can still move an activation beyond a bound, after which the indicator in Eq. 6 is zero for that neuron and the violation can persist. The paper states that the technique is equivalent to adding a clipping layer that 'acts as the identity in the forward pass', which means the constraint is never imposed on forward evaluations. The authors should either apply a forward clamp at inference time or empirically verify that final explanations satisfy Eq. 4, and they should soften the wording that claims the constraint 'ensures' the allowed activation range.
minor comments (4)
  1. [§5.1, Table 2] Table 2 reports no error bars, confidence intervals, or numbers of runs; the claim that FGVis outperforms the baselines by a large margin should be accompanied by variance estimates over image subsets or model seeds.
  2. [§5.1, Table 2] The reference-method results in Table 2 are taken from Petsiuk et al. [32] rather than rerun in the authors' pipeline. Please confirm that the experimental settings such as image preprocessing, deletion protocol, and model checkpoints are identical, or note any differences.
  3. [§4.3] The color-bias experiment uses small sample sizes (21 minivan and 42 school bus images); the 8.3% versus 83.3% comparison would benefit from a confidence interval or a statistical significance test.
  4. [A4.2] The supplementary material states that FGVis highlights mutations in the optic disk and blood vessels that are not covered by the ground-truth markings, producing visual false positives; this limitation should also be acknowledged in the main text's medical application section.

Circularity Check

1 steps flagged · score 6.0 of 10

Deletion-metric advantage largely restates the method's own loss; core defense is not circular.

  1. self definitional [Sec. 3.1 (Eq. 3); Sec. A2; Sec. 5.1 / A4.1]
    "Alternatively, we can compute a deleting explanation using: e*_cT = m*_cT * x, m*_cT = arg max_{m_cT} {phi(y^{cT}_x, y^{cT}_e) + lambda * ||m_cT||_1}. (3) ... For the similarity metric phi(.,.) we use the cross-entropy for the generation and preservation game and the negative probability for the deletion and repression game. ... The deletion metric is computed by measuring the area under the curve AUC of the deletion curve ... using the trapezoidal rule."

    The deletion game (Eq. 3) optimizes the mask to maximize the negative target-class softmax (plus sparsity), i.e. it searches for pixels whose removal lowers the target probability. The deletion metric (Sec. 5.1 / A4.1) removes exactly those pixels and scores the resulting drop. Thus the low AUC in Table 2 is the method's own objective, not an independent faithfulness test. Grad-CAM, LIME, RISE, and sliding-window occlusion do not optimize this objective, so the reported margin is structurally favored. Additionally, A4.1 uses a line-search for lambda that stops when y^{cT}_e < 0.02 * y^{cT}_x, placing the explanation at the bottom of the deletion curve before the metric is computed. The headline deletion-metric result is therefore substantially by construction.

full rationale

The central contribution, the gradient-clipping defense of Sec. 3.2, depends on an explicitly stated assumption (zero activations have the lowest information) and is not circular: the adversarial validations in Table 1 and Table A1 run an optimization game from zero/black images and empirically measure whether it can synthesize evidence; these checks do not reduce to the method's equations by construction. The clearest circularity is the faithfulness evaluation: Eq. 3's deletion game maximizes the negative target softmax, and the deletion metric removes the resulting mask and measures the target softmax drop. Since competing methods are not trained to minimize this same metric, the large deletion-metric advantage in Table 2 is partly guaranteed by construction. Moreover, A4.1's lambda line-search stops precisely when the explanation already satisfies a softmax below 2% of the original, so the explanation is already at the bottom of the deletion curve before the metric is evaluated. The medical lesion detection on DiaretDB1 uses expert-marked ground truth and is not circular. The citation to Gondal et al. [18] for the medical training setup is not load-bearing for the central claims. Concerns that the clip range of Eq. 4 permits deletion-based adversarial evidence are correctness risks, not circularity, and do not raise the score further.

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

The method introduces no new physical entities. It relies on a few standard assumptions about neural network optimization and on an explicit information-content assumption for zero activations. The free parameters are all optimization hyperparameters (λ, learning rate, binarization threshold), not fitted physical constants.

free parameters (3)
  • sparsity weight λ = line-search over 13 values between 1e-4 and 1e-10; fixed 1e-10 for medical
    Balances sparsity vs. faithfulness in Eq. 2-3; tuned per image via line-search, affecting all reported qualitative and quantitative results.
  • SGD learning rate = 0.1 default, 0.25 medical, 0.3 deletion metric
    Chosen by hand per experiment; affects convergence and mask quality, as shown in Fig. A2.
  • mask binarization threshold for medical evaluation = 4% of maximum mask value
    Converts continuous masks to binary for sensitivity reporting in Tab. 3; post-hoc threshold that changes the reported sensitivities.
assumptions (3)
  • domain assumption Zero activations carry the lowest information content.
    Underpins the clipping bounds in Eq. 4-5; if false, clipping could suppress informative features or admit adversarial ones. The paper cites network pruning [22] as justification.
  • standard math The CNN is differentiable with respect to the input and SGD can optimize the mask.
    Required for gradient-based optimization of Eq. 2-3 and for the gradient-clipping defense in Eq. 6.
  • domain assumption A zero image is an information-free reference that leads to high-entropy model output.
    Empirically justified in Sec. A3.1; used to define the removal operator Eq. 1 and the entire optimization-based explanation paradigm.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks." pith.science (2026). https://pith.science/paper/GAZU2ZSP

@misc{pith2026190802686,
  author       = {Pith},
  title        = {Pith review of: Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GAZU2ZSP}},
  note         = {Machine review of arXiv:1908.02686}
}
read the original abstract

To verify and validate networks, it is essential to gain insight into their decisions, limitations as well as possible shortcomings of training data. In this work, we propose a post-hoc, optimization based visual explanation method, which highlights the evidence in the input image for a specific prediction. Our approach is based on a novel technique to defend against adversarial evidence (i.e. faulty evidence due to artefacts) by filtering gradients during optimization. The defense does not depend on human-tuned parameters. It enables explanations which are both fine-grained and preserve the characteristics of images, such as edges and colors. The explanations are interpretable, suited for visualizing detailed evidence and can be tested as they are valid model inputs. We qualitatively and quantitatively evaluate our approach on a multitude of models and datasets.

Figures

Figures reproduced from arXiv: 1908.02686 by the authors.

Figure 1
Figure 1. Fine-grained explanations computed by remov [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visualization types calculated for VGG using deletion / preservation game. For the repression / generation game the same characteristics hold. Subscript cT ommited to ease readability. a) Input image. b) Mask obtained by the optimization. Colors in a deletion mask are complementary to the image colors. c) Explanation directly obtained by the optimization. d) Complementary mask with a true-color representation for th… view at source ↗
Figure 3
Figure 3. Explanations computed for the adversarial class [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Explanation masks for images with multiple ob [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Comparison of mean explanation masks: a) Image, b) BBMP [ [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Visual explanations computed using the deletion game for ResNet50. The masks (b, d) show a grid-like pat￾tern, as also observed in [31] for ResNet50. third row, are correctly discriminated. One major advantage of FGVis is its ability to visualize fine-grained details. …
Figure 7
Figure 7. Figure 7: Explanations computed using the preservation game for VGG16. Explanations of the class minivan focus on edges, hardly preserving the color, compared to the class school bus, with yellow dominating the explanations. fulnes of explanations generated by our method. This m…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 46 canonical work pages

  1. [1]

    Local explanation methods for deep neural networks lack sensitivity to parameter values

    Julius Adebayo, Justin Gilmer, Ian Goodfellow, and Been Kim. Local explanation methods for deep neural networks lack sensitivity to parameter values. In Workshop at the In- ternational Conference on Learning Representations (ICLR),

    work page

  2. [2]

    A survey on automated microaneurysm detection in dia- betic retinopathy retinal images

    Ankita Agrawal, Charul Bhatnagar, and Anand Singh Jalal. A survey on automated microaneurysm detection in dia- betic retinopathy retinal images. InInternational Conference on Information Systems and Computer Networks (ISCON) , pages 24–29. IEEE, 2013. 10

    work page 2013

  3. [3]

    Arunkumar and P

    R. Arunkumar and P. Karthigaikumar. Multi-retinal dis- ease classification by reduced deep learning features. Neural Computing and Applications, 28(2):329–334, 2017. 10

    work page 2017

  4. [4]

    On pixel-wise explanations for non-linear classi- fier decisions by layer-wise relevance propagation.PloS one, 10(7):e0130140, 2015

    Sebastian Bach, Alexander Binder, Gr ´egoire Montavon, Frederick Klauschen, Klaus-Robert M ¨uller, and Wojciech Samek. On pixel-wise explanations for non-linear classi- fier decisions by layer-wise relevance propagation.PloS one, 10(7):e0130140, 2015. 2

    work page 2015

  5. [5]

    How to explain individual classification decisions

    David Baehrens, Timon Schroeter, Stefan Harmeling, Mo- toaki Kawanabe, Katja Hansen, and Klaus-Robert M ¨uller. How to explain individual classification decisions. Journal of Machine Learning Research, 11(Jun):1803–1831, 2010. 2

    work page 2010

  6. [6]

    Explaining image classifiers by counter- factual generation

    Chun-Hao Chang, Elliot Creager, Anna Goldenberg, and David Duvenaud. Explaining image classifiers by counter- factual generation. arXiv e-prints, page arXiv:1807.08024, Jul 2018. 2, 3, 4, 5

    arXiv 2018

  7. [7]

    Balasubramanian

    Aditya Chattopadhay, Anirban Sarkar, Prantik Howlader, and Vineeth N. Balasubramanian. Grad-cam++: General- ized gradient-based visual explanations for deep convolu- tional networks. In Winter Conference on Applications of Computer Vision (WACV), pages 839–847, 2018. 2

    work page 2018

  8. [8]

    Colas, A

    E. Colas, A. Besse, A. Orgogozo, B. Schmauch, N. Meric, and E. Besse. Deep learning approach for diabetic retinopa- thy screening. Acta Ophthalmologica, 94(S256), 2016. 10

    work page 2016

Show all 54 references
  1. [9]

    The cityscapes dataset for semantic urban scene understanding

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Re...

    2016

  2. [10]

    EyePACS: an adapt- able telemedicine system for diabetic retinopathy screening

    Jorge Cuadros and George Bresnick. EyePACS: an adapt- able telemedicine system for diabetic retinopathy screening. Journal of Diabetes Science and Technology, 3(3):509–516,

  3. [11]

    Real time image saliency for black box classifiers

    Piotr Dabkowski and Yarin Gal. Real time image saliency for black box classifiers. In Advances in Neural Information Processing Systems (NIPS), pages 6967–6976, 2017. 1, 2, 3, 4, 5

    2017

  4. [12]

    Doll ´ar, C

    P. Doll ´ar, C. Wojek, B. Schiele, and P. Perona. Pedestrian de- tection: A benchmark. In Proceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition (CVPR) ,

  5. [13]

    Techniques for interpretable machine learning

    Mengnan Du, Ninghao Liu, and Xia Hu. Techniques for interpretable machine learning. arXiv e-prints , page arXiv:1808.00033, Jul 2018. 2

    2018 arXiv

  6. [14]

    To- wards explanation of dnn-based prediction with guided fea- ture inversion

    Mengnan Du, Ninghao Liu, Qingquan Song, and Xia Hu. To- wards explanation of dnn-based prediction with guided fea- ture inversion. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 1358–1367, 2018. 2, 3, 4

    2018

  7. [15]

    https://www.kaggle.com/c/diabetic-retinopathy- detection

    EyePACS. https://www.kaggle.com/c/diabetic-retinopathy- detection. assessed on 2018-09-23, 2015. 10

    2018

  8. [16]

    https://www.kaggle.com/c/diabetic-retinopathy- detection/discussion/15617

    EyePACS. https://www.kaggle.com/c/diabetic-retinopathy- detection/discussion/15617. assessed on 2018-09-23. 10

    2018

  9. [17]

    Fong and Andrea Vedaldi

    Ruth C. Fong and Andrea Vedaldi. Interpretable explanations of black boxes by meaningful perturbation. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pages 3429–3437, 2017. 1, 2, 3, 4, 5, 7

    2017

  10. [18]

    Gondal, Jan M

    Waleed M. Gondal, Jan M. K ¨ohler, Ren ´e Grzeszick, Ger- not A. Fink, and Michael Hirsch. Weakly-supervised local- ization of diabetic retinopathy lesions in retinal fundus im- ages. In IEEE International Conference on Image Process- ing (ICIP), pages 2069–2073, 2017. 1, 8, 10

    2017

  11. [19]

    Explaining and harnessing adversarial examples

    Ian Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. In Inter- national Conference on Learning Representations (ICLR) ,

  12. [20]

    Development and validation of a deep learning algo- rithm for detection of diabetic retinopathy in retinal fundus photographs

    Varun Gulshan, Lily Peng, Marc Coram, Martin C Stumpe, Derek Wu, Arunachalam Narayanaswamy, Subhashini Venu- gopalan, Kasumi Widner, Tom Madams, Jorge Cuadros, et al. Development and validation of a deep learning algo- rithm for detection of diabetic retinopathy in retinal fun...

    2016

  13. [21]

    A gaussian scale space approach for exudates detection, clas- sification and severity prediction

    Mrinal Haloi, Samarendra Dandapat, and Rohit Sinha. A gaussian scale space approach for exudates detection, clas- sification and severity prediction. arXiv e-prints , page arXiv:1505.00737, May 2015. 8

    2015 arXiv

  14. [22]

    Learning both weights and connections for efficient neural network

    Song Han, Jeff Pool, John Tran, and William Dally. Learning both weights and connections for efficient neural network. In Advances in Neural Information Processing Systems (NIPS), pages 1135–1143, 2015. 5

    2015

  15. [23]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 6

    2016

  16. [24]

    Distilling the knowledge in a neural network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv e-prints , page arXiv:1503.02531, Mar 2015. 3

    2015 arXiv

  17. [25]

    The DIARETDB1 diabetic retinopathy database and evaluation protocol

    Tomi Kauppi, Valentina Kalesnykiene, Joni-Kristian Ka- marainen, Lasse Lensu, Iiris Sorri, Asta Raninen, Raija V outilainen, Hannu Uusitalo, Heikki K¨alvi¨ainen, and Juhani Pietil¨a. The DIARETDB1 diabetic retinopathy database and evaluation protocol. In British Machine Vision...

    2007

  18. [26]

    ImageNet classification with deep convolutional neural net- works

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. ImageNet classification with deep convolutional neural net- works. In Advances in Neural Information Processing Sys- tems (NIPS), pages 1097–1105, 2012. 6, 1

    2012

  19. [27]

    Adver- sarial examples in the physical world

    Alexey Kurakin, Ian Goodfellow, and Samy Bengio. Adver- sarial examples in the physical world. arXiv e-prints, page arXiv:1607.02533, Jul 2016. 4, 5

    2016 arXiv

  20. [28]

    A location-to-segmentation strategy for automatic exudate segmentation in colour retinal fundus images

    Qing Liu, Beiji Zou, Jie Chen, Wei Ke, Kejuan Yue, Zail- iang Chen, and Guoying Zhao. A location-to-segmentation strategy for automatic exudate segmentation in colour retinal fundus images. Computerized Medical Imaging and Graph- ics, 55:78–86, 2017. 8

    2017

  21. [29]

    De- tection of red lesions in diabetic retinopathy affected fundus images

    Vijay M Mane, Ramish B Kawadiwale, and DV Jadhav. De- tection of red lesions in diabetic retinopathy affected fundus images. In IEEE International Advance Computing Confer- ence (IACC), pages 56–60, 2015. 8

    2015

  22. [30]

    Concrete problems for autonomous vehicle safety: Advantages of Bayesian deep learning

    Rowan McAllister, Yarin Gal, Alex Kendall, Mark Van Der Wilk, Amar Shah, Roberto Cipolla, and Adrian Vivian Weller. Concrete problems for autonomous vehicle safety: Advantages of Bayesian deep learning. InInternational Joint Conferences on Artificial Intelligence (IJCAI), 2017. 1

    2017

  23. [31]

    A theoretical ex- planation for perplexing behaviors of backpropagation-based visualizations

    Weili Nie, Yang Zhang, and Ankit Patel. A theoretical ex- planation for perplexing behaviors of backpropagation-based visualizations. arXiv e-prints, page arXiv:1805.07039, May

    arXiv

  24. [32]

    Rise: Random- ized input sampling for explanation of black-box models

    Vitali Petsiuk, Abir Das, and Kate Saenko. Rise: Random- ized input sampling for explanation of black-box models. In British Machine Vision Conference (BMVC), 2018. 1, 2, 7, 8, 10

    2018

  25. [33]

    Why should I trust you?: Explaining the predictions of any classifier

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. Why should I trust you?: Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD In- ternational Conference on Knowledge Discovery and Data Mining, pages 1135–1144, 2016. 2, 8

    2016

  26. [34]

    Berg, and Li Fei-Fei

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet large scale visual recognition chal- lenge. International Journal of Computer Vision ...

    2015

  27. [35]

    Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Ba- tra

    Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Ba- tra. Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE In- ternational Conference on Computer Vision (ICCV) , p...

    2017

  28. [36]

    Regional multi- scale approach for visually pleasing explanations of deep neural networks

    Dasom Seo, Kanghan Oh, and Il-Seok Oh. Regional multi- scale approach for visually pleasing explanations of deep neural networks. arXiv e-prints, page arXiv:1807.11720, Jul

    arXiv

  29. [37]

    Learning important features through propagating activation differences

    Avanti Shrikumar, Peyton Greenside, and Anshul Kundaje. Learning important features through propagating activation differences. In Proceedings of the 34th International Confer- ence on Machine Learning (ICML), pages 3145–3153, 2017. 2

    2017

  30. [38]

    Deep inside convolutional networks: Visualising image clas- sification models and saliency maps

    Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. Deep inside convolutional networks: Visualising image clas- sification models and saliency maps. International Confer- ence on Learning Representations (ICLR), 2014. 2, 7

    2014

  31. [39]

    Very deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition. arXiv e-prints, page arXiv:1409.1556, Sep 2014. 6

    2014 arXiv

  32. [40]

    Smoothgrad: removing noise by adding noise

    Daniel Smilkov, Nikhil Thorat, Been Kim, Fernanda Vi ´egas, and Martin Wattenberg. Smoothgrad: removing noise by adding noise. arXiv e-prints, page arXiv:1706.03825, Jun

    arXiv

  33. [41]

    Solomon, Emily Chew, Elia J

    Sharon D. Solomon, Emily Chew, Elia J. Duh, Lucia Sobrin, Jennifer K. Sun, Brian L. VanderBeek, Charles C. Wykoff, and Thomas W. Gardner. Diabetic retinopathy: a position statement by the American diabetes association. Diabetes care, 40(3):412–418, 2017. 10, 16

    2017

  34. [42]

    Striving for simplicity: The all convolutional net

    Jost Tobias Springenberg, Alexey Dosovitskiy, Thomas Brox, and Martin Riedmiller. Striving for simplicity: The all convolutional net. In International Conference on Learn- ing Representations (ICLR), 2015. 1, 2, 7

    2015

  35. [43]

    Axiomatic attribution for deep networks

    Mukund Sundararajan, Ankur Taly, and Qiqi Yan. Axiomatic attribution for deep networks. In Proceedings of the 34th In- ternational Conference on Machine Learning (ICML), pages 3319–3328, 2017. 2

    2017

  36. [44]

    Going deeper with convolutions

    Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015. 6

    2015

  37. [45]

    In- triguing properties of neural networks

    Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus. In- triguing properties of neural networks. In International Con- ference on Learning Representations (ICLR), 2014. 4

    2014

  38. [46]

    Development and validation of a deep learn- ing system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with dia- betes

    Daniel Shu Wei Ting, Carol Yim-Lui Cheung, Gilbert Lim, Gavin Siew Wei Tan, Nguyen D Quang, Alfred Gan, Haslina Hamzah, Renata Garcia-Franco, Ian Yew San Yeo, Shu Yen Lee, et al. Development and validation of a deep learn- ing system for diabetic retinopathy and related eye di...

    2017

  39. [47]

    Rogers, Ryo Kawasaki, Ecosse L

    Joanne WY Yau, Sophie L. Rogers, Ryo Kawasaki, Ecosse L. Lamoureux, Jonathan W. Kowalski, Toke Bek, Shih-Jen Chen, Jacqueline M. Dekker, Astrid Fletcher, Jakob Grauslund, et al. Global prevalence and major risk factors of diabetic retinopathy. Diabetes care, 35(3):556–564, 2012. 10

    2012

  40. [48]

    Zeiler and Rob Fergus

    Matthew D. Zeiler and Rob Fergus. Visualizing and under- standing convolutional networks. In Proceedings of the Eu- ropean Conference on Computer Vision (ECCV), pages 818– 833, 2014. 2, 7, 8

    2014

  41. [49]

    Top-down neural attention by excitation backprop

    Jianming Zhang, Zhe Lin, Jonathan Brandt, Xiaohui Shen, and Stan Sclaroff. Top-down neural attention by excitation backprop. In Proceedings of the European Conference on Computer Vision (ECCV), pages 543–559, 2016. 1, 2, 7

    2016

  42. [50]

    Visual interpretability for deep learning: A survey

    Quan-shi Zhang and Song-Chun Zhu. Visual interpretability for deep learning: A survey. Frontiers of Information Tech- nology & Electronic Engineering, 19(1):27–39, 2018. 2

    2018

  43. [51]

    Uniqueness-driven saliency analysis for automated lesion detection with applications to retinal diseases

    Yitian Zhao, Yalin Zheng, Yifan Zhao, Yonghuai Liu, Zhili Chen, Peng Liu, and Jiang Liu. Uniqueness-driven saliency analysis for automated lesion detection with applications to retinal diseases. In International Conference on Medical Im- age Computing and Computer-Assisted Int...

    2018

  44. [52]

    Object Detectors Emerge in Deep Scene CNNs

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Object Detectors Emerge in Deep Scene CNNs. arXiv e-prints , page arXiv:1412.6856, Dec

    arXiv

  45. [53]

    Learning deep features for discrimi- native localization

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Learning deep features for discrimi- native localization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2921–2929, 2016. 1, 2, 8

    2016

  46. [54]

    Au- tomatic hemorrhage detection in color fundus images based on gradual removal of vascular branches

    Lei Zhou, Penglin Li, Qi Yu, Yu Qiao, and Jie Yang. Au- tomatic hemorrhage detection in color fundus images based on gradual removal of vascular branches. In IEEE Interna- tional Conference on Image Processing (ICIP) , pages 399– 403, 2016. 8 Interpretable and Fine-Grained Vis...

    2016

Pith tools

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