REAPS: Towards Better Recognition of Fine-grained Images by Region Attending and Part Sequencing

Peng Zhang; Shuigeng Zhou; Xinyu Zhu; Yi Niu; Zhanzhan Cheng

REVIEW 4 major objections 6 minor 40 references

REAPS: Towards Better Recognition of Fine-grained Images by Region Attending and Part Sequencing

T0 review · 4 major / 6 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 A single network, trained with only image-level labels, reaches state-of-the-art fine-grained recognition by attending to the object region and modeling its parts as a sequence.

desk verdict REAPS is a plausible incremental combination of CAM cropping and sequence learning, but its state-of-the-art claim does not survive a direct comparison with the omitted NTS-Net baseline. read the letter →

arxiv 1908.01962 v1 pith:SFHST6I6 submitted 2019-08-06 cs.CV

Peng Zhang , Xinyu Zhu , Zhanzhan Cheng , Shuigeng Zhou , Yi Niu This is my paper · ORCID

classification cs.CV

keywords fine-grainedimagerecognitionweaklysupervisedlearningclassactivationmappingregionattentionpartsequencebi-directionalLSTMsoftpartsREAPS

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

REAPS is a unified network for fine-grained image recognition that needs only image-level category labels. It first uses class activation mapping to locate and amplify the object region, suppressing background, and then turns the region's feature map into a sequence of local descriptors that a bi-directional LSTM maps to implicit 'soft parts,' capturing spatial relationships among parts without fixing their number. Trained end-to-end on three benchmarks, it reports state-of-the-art accuracies of 93.5% on Stanford Cars, 92.6% on FGVC-Aircraft, and 86.8% on CUB-200-2011, beating methods that use bounding boxes or part annotations. The paper's claim is that part-based fine-grained recognition can be done more simply and more flexibly by replacing explicit part detectors with sequence modeling on an attended crop.

What carries the argument

The load-bearing mechanism is the 'soft-part' sequence model. After the CAM-based region attending network crops and zooms the attended region, the backbone's feature map $X\in\mathbb{R}^{H\times W\times C}$ is decomposed by a pooling operation into a sequence $Y=[Y_1,\dots,Y_N]$ of $C$-dimensional vectors, each describing a rectangular region of the raw image. A bi-directional LSTM then maps this sequence into $M$ implicit part representations, and the concatenated hidden states serve as the part-level feature; this replaces hard part localization with a learned ordering over spatial descriptors. The second component, the class activation map from Eq. (1), is what suppresses background and defines the attended region that the sequence model sees.

What would settle it

A concrete check: on CUB-200-2011, compute the CAM bounding box at $\tau=0.1$ for every test image and measure how often ground-truth part keypoints such as head, wing, or tail fall outside the crop; then retrain the pipeline under several values of $\tau$. If a substantial fraction of diagnostic parts fall outside at $\tau=0.1$, or if accuracy increases when the crop is enlarged, the central premise that the attended region preserves all discriminative details is disproved.

Watch

Extended reading notes

Core claim

The central discovery is that two previously separate ideas, region-level attention and part-level modeling, can be fused into one end-to-end, weakly supervised pipeline that matches or exceeds stronger-supervised methods. On Stanford Cars, FGVC-Aircraft, and CUB-200-2011, the proposed REAPS+ reaches 93.5%, 92.6%, and 86.8% respectively. The largest margin over the prior state of the art appears on FGVC-Aircraft, where REAPS+ surpasses MA-CNN by a relative 2.7% gain; the authors attribute this to the sequence model exploiting the strong spatial and sequential structure of aircraft. The paper also reports that adding the part-sequence branch to a plain backbone gives a relative performance gain of about 1.0% on Stanford Cars, and that the region-attending component outperforms FCAN and RA-CNN in the single-scale comparison.

Load-bearing premise

The method's accuracy rests on the assumption that the CAM-based attended region, cropped with threshold $\tau=0.1$, contains every detail needed to tell the classes apart; if the crop leaves out a diagnostic part, the part-sequence branch never gets to see it.

Editorial extensions

If this is right

If REAPS is right, fine-grained recognition no longer needs bounding boxes or part keypoints at any stage; image-level labels are enough, which lowers annotation cost dramatically.
Because parts are implicit and their count is not fixed, the method should tolerate occlusion and incomplete objects better than approaches that detect a preset number of semantic parts.
Modeling parts as an ordered sequence lets the network exploit spatial context among parts, so categories with strong geometric ordering, such as aircraft, cars, and birds, stand to benefit most.
The framework is stackable: feeding the attended region from one PSN into a second PSN yields further gains of 93.5%, 92.6%, and 86.8%, suggesting that accuracy improves with depth of region refinement.

Reading between the lines

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

My inference: a testable extension the paper leaves open is to vary the CAM crop threshold $\tau$ and measure accuracy; if performance is sensitive to $\tau$, the method's robustness claim depends on a single unablated hyperparameter.
My inference: the same sequence mapping could be applied to other fine-grained domains where parts are naturally ordered, such as text lines, handwritten characters, or fingerprint ridges.
My inference: the paper's results on FGVC-Aircraft suggest, though the authors do not test it explicitly, that the sequence model's benefit grows with the regularity of part order, so categories with highly variable part arrangements might show smaller gains.

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

The paper is a competently written, plausible piece of incremental work. The new bit is the 'soft-part' idea: instead of detecting a fixed number of parts, serialize the spatial feature vectors and run a bidirectional LSTM to learn implicit part representations. That is a neat framing and the end-to-end training with three losses is sensible. The experiments cover the three standard fine-grained datasets and the reported numbers are competitive with several baselines. The paper clearly explains the motivation and the method, and the ablation showing the part branch helps is useful.

The problems are in the comparison and the claims. Table 3 omits NTS-Net (ECCV 2018), which is directly relevant and was published before this work. NTS-Net reports 93.9% on Stanford Cars and 87.5% on CUB, both above REAPS+ (93.5 and 86.8). The stress-test note is right: the abstract and Section 4.2 say 'state of the art', and on two of three benchmarks that is simply false under the published comparison. The text also says REAPS+ surpasses MA-CNN by a relative 2.7% on Aircraft, but on CUB REAPS (86.0) is below MA-CNN (86.5), so the overall superiority claim is undercut. This is not a matter of heavy hedging; it is a factual gap that any careful reader can check.

Second, the paper never specifies the sequence length N, the implicit part count M, or the LSTM hidden size U. Those are core hyperparameters for reproducing the method. The CAM threshold tau is set to 0.1 but there is no ablation on it, so the concern that the attended region may drop discriminative details is real but unquantified. No error bars are given, which is common but still a limitation when the margins are under one percent. The claim that REAPS is simple and robust is therefore not fully substantiated.

The idea itself is legitimately a combination of known ingredients, and the paper does not hide that; it cites CAM and RA-CNN. The novelty is in the combination, not in the components. That is enough for a workshop or a solid conference paper with careful revision.

Who should read this? People working on weakly supervised fine-grained recognition will find the sequence-modeling approach worth knowing, especially the ablations. It should go to peer review, not be desk-rejected, but it needs a major revision: add NTS-Net, specify all hyperparameters, report error bars if possible, and tone down the SOTA claim to 'competitive'. As it stands, the evidence supports the method working reasonably well, not setting a new state of the art.

Referee Report

4 major / 6 minor

Summary. The paper proposes REAPS, a weakly supervised fine-grained recognition framework that combines a CAM-based region attending network (RAN) with a part sequence-learning network (PSN). RAN computes a class activation map (Eq. 1), thresholds it to crop and zoom in on an attended region (Eq. 2), and PSN serializes the backbone features into N vectors, maps them to M implicit part representations with a bi-directional LSTM, and classifies with three combined softmax losses (Eq. 5). Experiments on Stanford Cars, FGVC-Aircraft, and CUB-200-2011 report REAPS+ accuracies of 93.5%, 92.6%, and 86.8%, respectively, and the paper claims state-of-the-art performance with only image-level labels.

Significance. The conceptual contribution is attractive: replacing fixed hard part detectors with an implicit soft-part sequence model on a CAM-cropped object region is simple, end-to-end trainable, and requires no part annotations. The paper evaluates on three standard benchmarks with a consistent backbone, and the qualitative visualizations support the claim that the part branch produces more compact feature maps. However, the central state-of-the-art claim is not supported by the published comparison: the directly comparable weakly supervised method NTS-Net (ECCV 2018) reports 93.9% on Stanford Cars, 91.4% on FGVC-Aircraft, and 87.5% on CUB-200-2011 with image-level labels, so REAPS+ is higher only on Aircraft. In addition, key hyperparameters needed for reproduction are not specified. The paper contains no machine-checked proofs or released code, so its value rests on the completeness and correctness of the empirical comparison, and that comparison is currently incomplete.

major comments (4)

[Table 3 / Sec. 4.2] Table 3 and the discussion in Section 4.2 omit NTS-Net (Yang et al., ECCV 2018), a directly comparable weakly supervised fine-grained recognition method that reports 93.9% on Stanford Cars, 91.4% on FGVC-Aircraft, and 87.5% on CUB-200-2011 using only image-level labels. Under this published comparison, REAPS+ (93.5%, 92.6%, 86.8%) is not state of the art on two of the three benchmarks, and the sentence in Section 4.2 that REAPS+ "obtains the best performance on three datasets" is not supportable. The authors must add NTS-Net and any other comparable published weakly supervised baselines, and revise the state-of-the-art claims accordingly.
[Sec. 3.3 / Sec. 4.1] Key hyperparameters of the proposed architecture are never specified: the sequence length N in Eq. (3), the implicit part count M in Eq. (4), the LSTM hidden size U, the number of LSTM layers, and the scaling factor used for the second PSN in REAPS+. Section 4.1 states only the loss weights lambda_1=lambda_2=lambda_3=1, the threshold tau=0.1, and the learning-rate schedule. Without these values the experiments cannot be reproduced, and the sensitivity of the reported margins to these choices cannot be assessed.
[Eq. (2) / Sec. 4.2] The central design assumption is that the thresholded CAM crop in Eq. (2) preserves all diagnostic fine-grained parts before the part sequence model sees them, but the paper does not ablate tau or quantify the information loss caused by cropping. If the class activation map highlights only the most salient patch and drops other diagnostic regions, the crop removes information before PSN can model it; an ablation over tau values and a comparison of REAPS with and without cropping would be needed to substantiate the claim in Section 4.2 that RAN "effectively depress[es] background noise" without sacrificing fine-grained details.
[Table 3 / Sec. 4.2] The reported differences are small on two of the three benchmarks (93.5 vs. 92.8 for MA-CNN and 93.9 for NTS-Net on Cars; 86.8 vs. 86.5 for MA-CNN and 87.5 for NTS-Net on CUB), yet no standard deviations, confidence intervals, or significance tests are reported. Given the missing baseline and unspecified hyperparameters, the conclusion that REAPS "outperforms the existing methods" is not robust; at minimum the authors should report the variance over multiple runs or justify single-run reporting.

minor comments (6)

[Sec. 4.2] The statement that REAPS+ surpasses MA-CNN "by a relative 2.7% gain" on FGVC-Aircraft is arithmetically inaccurate: 92.6% vs. 89.9% is a 2.7 percentage-point improvement and a 3.0% relative improvement.
[Sec. 4.2 / Table 2] The text says the part branch gives a "relative performance gain of 1.0%" for PSN on Stanford Cars, but the numbers 91.3% and 92.3% differ by 1.0 percentage point; the wording should be corrected.
[Sec. 3.3 / Eq. (4)] There is an apparent inconsistency between the notation in Eq. (4), where the mapped part representation is [P_1,...,P_M] with M <= N, and the later sentence stating that the concatenated LSTM hidden states give P_P in R^{N x U}; the authors should clarify whether M=N in their implementation or describe the additional reduction step.
[Table 3] The baseline named "REAPS wo PSN" is not clearly defined in the text; it should be stated explicitly which branches are removed and how this differs from the "PSN wo part" model in Table 2.
[Abstract / Sec. 5] The abstract and conclusion claim "state of the art performance" without qualification; given the omitted NTS-Net comparison, this wording should be revised to a more defensible claim such as "competitive with current methods" unless the comparison is updated.
[Sec. 3.3 / Fig. 3] The notation U and M appears in Fig. 3 without definition in the caption or main text; please define all symbols at first use.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: REAPS is an empirical pipeline with fixed hyperparameters and external baselines.

full rationale

The paper's pipeline (CAM-based region attending, cropping, part-sequence LSTM, multi-loss training) contains no step in which an output quantity is defined in terms of itself, nor any 'prediction' that reduces to a fitted parameter. The only thresholds and weights (tau=0.1, lambda_j=1) are fixed manually and not tuned on the test set. The attention map in Eq. (1) is a standard weighted sum of convolutional activations using softmax weights; the crop in Eq. (2) uses a fixed threshold; the sequence mapping in Eq. (4) is a learned bi-LSTM. Accuracy is reported on held-out test sets from standard benchmarks. All cited prior works are external; there are no self-citations or imported uniqueness theorems that carry the argument. The omission of NTS-Net from Table 3 is a potential correctness/comparison gap, not a circularity. Therefore the derivation chain is self-contained and non-circular.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The empirical claims rest on standard transfer-learning assumptions and on two unverified modeling assumptions: CAM crops are object-complete, and serialized block pooling preserves discriminative detail. The paper also leaves central sequence hyperparameters unspecified, so the method as published is under-specified.

free parameters (5)

CAM bounding-box threshold tau = 0.1
Set by hand in Eq. (2); controls which pixels are kept in the attended region and is never ablated.
sequence length N = not reported
Defines how the feature map is decomposed in Eq. (3); central to the part sequence model but never specified.
implicit part count M = not reported
Number of implicit parts produced by the mapping in Eq. (4); affects LSTM output size and is never given.
LSTM hidden size U = not reported
Determines capacity of the part representation in Section 3.3; not reported.
loss weights lambda_1, lambda_2, lambda_3 = 1, 1, 1
Set equally in Section 4.1; no sensitivity analysis is provided.

assumptions (4)

domain assumption ImageNet-pretrained VGG19 features transfer to the three fine-grained datasets, and fine-tuning the same backbone is sufficient for fair comparison.
Invoked throughout Section 4.1 for fair comparison; no analysis of backbone sensitivity is given.
domain assumption The CAM-based region attention from Eq. (1), thresholded at tau, localizes the full object region rather than only a few discriminative spots.
Load-bearing for RAN; if the crop misses parts, PSN cannot recover them (Section 3.2).
domain assumption Serializing the feature map into N rectangular blocks and pooling each block preserves the fine-grained visual details needed for classification.
Central to Eq. (3); average pooling over large blocks could discard subtle part differences.
domain assumption A bidirectional LSTM can map the source sequence of N feature vectors to M implicit part representations without alignment or part-level supervision.
Stated in Section 3.3; the paper provides no evidence that the hidden states correspond to interpretable parts.

how reviews work

0 comments

Cite this review

Pith. "Pith review of REAPS: Towards Better Recognition of Fine-grained Images by Region Attending and Part Sequencing." pith.science (2026). https://pith.science/paper/SFHST6I6

@misc{pith2026190801962,
  author       = {Pith},
  title        = {Pith review of: REAPS: Towards Better Recognition of Fine-grained Images by Region Attending and Part Sequencing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SFHST6I6}},
  note         = {Machine review of arXiv:1908.01962}
}

read the original abstract

Fine-grained image recognition has been a hot research topic in computer vision due to its various applications. The-state-of-the-art is the part/region-based approaches that first localize discriminative parts/regions, and then learn their fine-grained features. However, these approaches have some inherent drawbacks: 1) the discriminative feature representation of an object is prone to be disturbed by complicated background; 2) it is unreasonable and inflexible to fix the number of salient parts, because the intended parts may be unavailable under certain circumstances due to occlusion or incompleteness, and 3) the spatial correlation among different salient parts has not been thoroughly exploited (if not completely neglected). To overcome these drawbacks, in this paper we propose a new, simple yet robust method by building part sequence model on the attended object region. Concretely, we first try to alleviate the background effect by using a region attention mechanism to generate the attended region from the original image. Then, instead of localizing different salient parts and extracting their features separately, we learn the part representation implicitly by applying a mapping function on the serialized features of the object. Finally, we combine the region attending network and the part sequence learning network into a unified framework that can be trained end-to-end with only image-level labels. Our extensive experiments on three fine-grained benchmarks show that the proposed method achieves the state of the art performance.

Figures

Figures reproduced from arXiv: 1908.01962 by the authors.

**Figure 1.** An illustrative comparison between our framework and two popular existing fine-grained recognition frameworks. (a), (b) and (c) represent RLA, PL and our proposed framework respectively. O, Pi, Ri and Bi correspond to the original image, a selected part, an attended region and a backbone network respectively. + represents the operation of crop and zoom in. These drawbacks mentioned above motivate us to develop a ne… view at source ↗

**Figure 2.** Drawbacks of the RLA and PL frameworks. (a) RLA focuses on a detailed region progressively, while neglecting some other salient parts (the feet and wings of the bird disappear in scale1 and scale2 images). (b) PL detects a fixed number of preset parts and may get into trouble when some intended parts miss or be occluded (the back of the bird cannot be seen). [Best viewed in color] region attending network for gener… view at source ↗

**Figure 3.** The REAPS architecture. The region attending network (RAN) takes an original image as input and produces the region attention by weighting the last convolutional feature maps with the parameters of softmax layer. The attended region is cropped out and zoomed in, and fed into the part sequence learning network (PSN) where part representation is learned in an implicit way by applying the mapping function on the ser… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Region attention localization results of RAN for some examples from (a) Stanford Cars, (b) FGVC-aircraft, and (c) CUB-200-2011. Pictures from left to right in (a-c) are the raw image, the attention mask with bounding box indicating the area of top attention response, …

**Figure 5.** Figure 5: Visualization of feature maps for some examples from (a) Stanford Cars, (b) FGVC Aircraft, (c) CUB-200-2011. Pictures from left to right in (a-c) are the raw image, the feature map generated by PSN without part branch and the feature map generated by PSN with part bran…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 38 canonical work pages

[1]

In: CVPR

Angelova, A., Zhu, S.: Eﬃcient Object Detection and Segmentation for Fine- Grained Recognition. In: CVPR. pp. 811–818 (2013)

work page 2013
[2]

In: CVPR

Berg, T., Belhumeur, P.N.: POOF: Part-Based One-vs.-One Features for Fine- Grained Categorization, Face Veriﬁcation, and Attribute Estimation. In: CVPR. pp. 955–962 (2013)

work page 2013
[3]

In: CVPR

Berg, T., Liu, J., Lee, S.W., Alexander, M.L., Jacobs, D.W., Belhumeur, P.N.: Birdsnap: Large-Scale Fine-Grained Visual Categorization of Birds. In: CVPR. pp. 2019–2026 (2014)

work page 2014
[4]

CoRR abs/1406.2952 (2014)

Branson, S., Horn, G.V., Belongie, S.J., Perona, P.: Bird Species Categorization Using Pose Normalized Deep Convolutional Nets. CoRR abs/1406.2952 (2014)

arXiv 2014
[5]

In: CVPR

Cai, S., Zuo, W., Zhang, L.: Higher-Order Integration of Hierarchical Convolutional Activations for Fine-Grained Visual Categorization. In: CVPR. pp. 511–520 (2017)

work page 2017
[6]

In: EMNLP

Cho, K., van Merrienboer, B., Bahdanau, D., Bengio, Y.: On the Properties of Neural Machine Translation: Encoder-Decoder Approaches. In: EMNLP. pp. 103– 111 (2014)

work page 2014
[7]

In: EMNLP

Cho, K., van Merrienboer, B., G¨ ul¸ cehre, C ¸ ., Bahdanau, D., Bougares, F., Schwenk, H., Bengio, Y.: Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. In: EMNLP. pp. 1724–1734 (2014)

work page 2014
[8]

In: CVPR

Cui, Y., Zhou, F., Wang, J., Liu, X., Lin, Y., Belongie, S.J.: Kernel Pooling for Convolutional Neural Networks. In: CVPR. pp. 3049–3058 (2017)

work page 2017

Show all 40 references

[9]

In: CVPR

Fu, J., Zheng, H., Mei, T.: Look Closer to See Better: Recurrent Attention Con- volutional Neural Network for Fine-Grained Image Recognition. In: CVPR. pp. 4476–4484 (2017)

2017
[10]

In: CVPR

Gao, Y., Beijbom, O., Zhang, N., Darrell, T.: Compact Bilinear Pooling. In: CVPR. pp. 317–326 (2016)

2016
[11]

In: ICCV

Gavves, E., Fernando, B., Snoek, C.G.M., Smeulders, A.W.M., Tuytelaars, T.: Fine-Grained Categorization by Alignments. In: ICCV. pp. 1713–1720 (2013)

2013
[12]

IEEE Trans

He, X., Peng, Y., Zhao, J.: Fast Fine-Grained Image Classiﬁcation via Weakly Supervised Discriminative Localization. IEEE Trans. Circuits Syst. Video Techn. 29(5), 1394–1407 (2019)

2019
[13]

Neural Computation 9(8), 1735–1780 (1997)

Hochreiter, S., Schmidhuber, J.: Long Short-Term Memory. Neural Computation 9(8), 1735–1780 (1997)

1997
[14]

In: CVPR

Huang, S., Xu, Z., Tao, D., Zhang, Y.: Part-Stacked CNN for Fine-Grained Visual Categorization. In: CVPR. pp. 1173–1182 (2016)

2016
[15]

In: NIPS

Jaderberg, M., Simonyan, K., Zisserman, A., Kavukcuoglu, K.: Spatial Transformer Networks. In: NIPS. pp. 2017–2025 (2015)

2015
[16]

In: CVPR Workshop on FGVC

Khosla, A., Jayadevaprakash, N., Yao, B., Li, F.F.: Novel dataset for ﬁne-grained image categorization: Stanford dogs. In: CVPR Workshop on FGVC. vol. 2 (2011)

2011
[17]

In: CVPR

Kong, S., Fowlkes, C.C.: Low-Rank Bilinear Pooling for Fine-Grained Classiﬁca- tion. In: CVPR. pp. 7025–7034 (2017)

2017
[18]

In: CVPR

Krause, J., Jin, H., Yang, J., Li, F.: Fine-grained recognition without part anno- tations. In: CVPR. pp. 5546–5555 (2015)

2015
[19]

In: CVPR

Lin, D., Shen, X., Lu, C., Jia, J.: Deep LAC: deep localization, alignment and classiﬁcation for ﬁne-grained recognition. In: CVPR. pp. 1666–1674 (2015)

2015
[20]

In: BMVC (2017)

Lin, T., Maji, S.: Improved Bilinear Pooling with CNNs. In: BMVC (2017)

2017
[21]

In: ICCV

Lin, T., Roy Chowdhury, A., Maji, S.: Bilinear CNN Models for Fine-Grained Visual Recognition. In: ICCV. pp. 1449–1457 (2015) 12 Peng et al

2015
[22]

In: ECCV

Liu, J., Kanazawa, A., Jacobs, D.W., Belhumeur, P.N.: Dog Breed Classiﬁcation Using Part Localization. In: ECCV. pp. 172–185 (2012)

2012
[23]

CoRR abs/1603.06765 (2016)

Liu, X., Xia, T., Wang, J., Lin, Y.: Fully Convolutional Attention Localization Networks: Eﬃcient Attention Localization for Fine-Grained Recognition. CoRR abs/1603.06765 (2016)

2016 arXiv
[24]

HAL - INRIA (2013)

Maji, S., Rahtu, E., Kannala, J., Blaschko, M.B., Vedaldi, A.: Fine-Grained Visual Classiﬁcation of Aircraft. HAL - INRIA (2013)

2013
[25]

In: BMVC (2016)

Moghimi, M., Belongie, S.J., Saberian, M.J., Yang, J., Vasconcelos, N., Li, L.: Boosted Convolutional Neural Networks. In: BMVC (2016)

2016
[26]

In: Sixth Indian Conference on Computer Vision, Graphics & Image Processing, Bhubaneswar, India

Nilsback, M., Zisserman, A.: Automated Flower Classiﬁcation over a Large Number of Classes. In: Sixth Indian Conference on Computer Vision, Graphics & Image Processing, Bhubaneswar, India. pp. 722–729 (2008)

2008
[27]

In: ECCV

Perronnin, F., S´ anchez, J., Mensink, T.: Improving the Fisher Kernel for Large- Scale Image Classiﬁcation. In: ECCV. pp. 143–156 (2010)

2010
[28]

In: CVPR

Qian, Q., Jin, R., Zhu, S., Lin, Y.: Fine-grained visual categorization via multi- stage metric learning. In: CVPR. pp. 3716–3724 (2015)

2015
[29]

Wah, C., Branson, S., Welinder, P., Perona, P., Belongie, S.: The Caltech-UCSD Birds-200-2011 Dataset. Tech. Rep. CNS-TR-2011-001, California Institute of Technology (2011)

2011
[30]

In: ICCV

Wang, D., Shen, Z., Shao, J., Zhang, W., Xue, X., Zhang, Z.: Multiple Granularity Descriptors for Fine-Grained Categorization. In: ICCV. pp. 2399–2406 (2015)

2015
[31]

In: CVPR

Wang, Y., Choi, J., Morariu, V.I., Davis, L.S.: Mining Discriminative Triplets of Patches for Fine-Grained Classiﬁcation. In: CVPR. pp. 1163–1172 (2016)

2016
[32]

Pattern Recognition 76, 704–714 (2018)

Wei, X., Xie, C., Wu, J., Shen, C.: Mask-CNN: Localizing parts and selecting descriptors for ﬁne-grained bird species categorization. Pattern Recognition 76, 704–714 (2018)

2018
[33]

In: CVPR

Xiao, T., Xu, Y., Yang, K., Zhang, J., Peng, Y., Zhang, Z.: The application of two-level attention models in deep convolutional neural network for ﬁne-grained image classiﬁcation. In: CVPR. pp. 842–850 (2015)

2015
[34]

In: ICML

Xu, K., Ba, J., Kiros, R., Cho, K., Courville, A.C., Salakhutdinov, R., Zemel, R.S., Bengio, Y.: Show, Attend and Tell: Neural Image Caption Generation with Visual Attention. In: ICML. pp. 2048–2057 (2015)

2015
[35]

In: CVPR

Zhang, N., Farrell, R., Darrell, T.: Pose pooling kernels for sub-category recogni- tion. In: CVPR. pp. 3665–3672 (2012)

2012
[36]

In: ICCV

Zhang, N., Farrell, R., Iandola, F.N., Darrell, T.: Deformable Part Descriptors for Fine-Grained Recognition and Attribute Prediction. In: ICCV. pp. 729–736 (2013)

2013
[37]

In: CVPR

Zhang, X., Zhou, F., Lin, Y., Zhang, S.: Embedding Label Structures for Fine- Grained Feature Representation. In: CVPR. pp. 1114–1123 (2016)

2016
[38]

In: CVPR

Zhang, X., Xiong, H., Zhou, W., Lin, W., Tian, Q.: Picking Deep Filter Responses for Fine-Grained Image Recognition. In: CVPR. pp. 1134–1142 (2016)

2016
[39]

In: ICCV

Zheng, H., Fu, J., Mei, T., Luo, J.: Learning Multi-attention Convolutional Neural Network for Fine-Grained Image Recognition. In: ICCV. pp. 5219–5227 (2017)

2017
[40]

In: CVPR

Zhou, B., Khosla, A., Lapedriza, `A., Oliva, A., Torralba, A.: Learning Deep Fea- tures for Discriminative Localization. In: CVPR. pp. 2921–2929 (2016)

2016

Pith tools

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

[1] [1]

In: CVPR

Angelova, A., Zhu, S.: Eﬃcient Object Detection and Segmentation for Fine- Grained Recognition. In: CVPR. pp. 811–818 (2013)

work page 2013

[2] [2]

In: CVPR

Berg, T., Belhumeur, P.N.: POOF: Part-Based One-vs.-One Features for Fine- Grained Categorization, Face Veriﬁcation, and Attribute Estimation. In: CVPR. pp. 955–962 (2013)

work page 2013

[3] [3]

In: CVPR

Berg, T., Liu, J., Lee, S.W., Alexander, M.L., Jacobs, D.W., Belhumeur, P.N.: Birdsnap: Large-Scale Fine-Grained Visual Categorization of Birds. In: CVPR. pp. 2019–2026 (2014)

work page 2014

[4] [4]

CoRR abs/1406.2952 (2014)

Branson, S., Horn, G.V., Belongie, S.J., Perona, P.: Bird Species Categorization Using Pose Normalized Deep Convolutional Nets. CoRR abs/1406.2952 (2014)

arXiv 2014

[5] [5]

In: CVPR

Cai, S., Zuo, W., Zhang, L.: Higher-Order Integration of Hierarchical Convolutional Activations for Fine-Grained Visual Categorization. In: CVPR. pp. 511–520 (2017)

work page 2017

[6] [6]

In: EMNLP

Cho, K., van Merrienboer, B., Bahdanau, D., Bengio, Y.: On the Properties of Neural Machine Translation: Encoder-Decoder Approaches. In: EMNLP. pp. 103– 111 (2014)

work page 2014

[7] [7]

In: EMNLP

Cho, K., van Merrienboer, B., G¨ ul¸ cehre, C ¸ ., Bahdanau, D., Bougares, F., Schwenk, H., Bengio, Y.: Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. In: EMNLP. pp. 1724–1734 (2014)

work page 2014

[8] [8]

In: CVPR

Cui, Y., Zhou, F., Wang, J., Liu, X., Lin, Y., Belongie, S.J.: Kernel Pooling for Convolutional Neural Networks. In: CVPR. pp. 3049–3058 (2017)

work page 2017