Pith. sign in

REVIEW 3 major objections 5 minor 47 references

Decoupled Box Proposal and Featurization with Ultrafine-Grained Semantic Labels Improve Image Captioning and Visual Question Answering

T0 review · 3 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 Ultrafine-grained box embeddings beat object-detector features for captioning and VQA.

desk verdict A clean empirical comparison that makes a plausible case for ultrafine-grained box features, but the missing overlap analysis between Graph-RISE pretraining data and Conceptual Captions leaves the headline gain partly unsecured. read the letter →

arxiv 1909.02097 v1 pith:DIJL72J3 submitted 2019-09-04 cs.CL cs.CV

Soravit Changpinyo , Bo Pang , Piyush Sharma , Radu Soricut This is my paper · ORCID
classification cs.CLcs.CV
keywords imagecaptioningvisualquestionansweringobjectdetectiontransferlearningultrafine-grainedsemanticlabelsGraph-RISEdecoupledboxproposalConceptualCaptions
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

The paper argues that the two halves of object detection—proposing boxes and featurizing them—can be separated, and that improving featurization alone improves downstream vision-and-language tasks. It keeps Faster R-CNN's box proposals but swaps the 2048-dimensional detector features for 64-dimensional embeddings from Graph-RISE, a model trained on roughly 260 million web images with noisy, ultrafine-grained semantic labels. On Conceptual Captions, this raises CIDEr from 0.944 to 0.984; on VizWiz, VQA accuracy rises from 51.9% to 53.7%. If right, this means expensive box-and-label annotation is not the only route to effective bottom-up features, and fine-grained label granularity matters for transfer to out-of-domain images.

What carries the argument

The load-bearing object is Graph-RISE's 64-dimensional image-semantic embedding, trained on roughly 260 million web images with about 40 million noisy semantic labels, and organized so that semantic similarity is preserved from coarse category (bridge) down to instance (golden gate bridge). In the decoupled pipeline, Faster R-CNN's Region Proposal Network still supplies cropped boxes, and each crop is either featurized by Faster R-CNN's 2048-dimensional fc6 features or by Graph-RISE's 64-dimensional embedding; the captioning model adds a third channel of label embeddings from a word2vec model. The argument isolates the featurizer by holding boxes fixed, so any score difference is attributed to the featurization and its label granularity.

What would settle it

Retrain or filter Graph-RISE so that none of its roughly 260 million web training images overlap with Conceptual Captions or VizWiz, then rerun the captioning and VQA comparisons; if the CIDEr and accuracy advantages disappear, the claimed transfer effect is explained by pretraining-data overlap instead of ultrafine-grained labels.

Watch

Extended reading notes

Core claim

The paper's central claim is that decoupling box proposal from featurization lets downstream tasks use a much richer featurizer than an object detector trained on Visual Genome, and that ultrafine-grained semantic labels (instance-level rather than category-level) are what make the transfer work. The authors support this with controlled comparisons that hold the box proposals, architecture, and training recipe fixed, changing only whether box features come from Faster R-CNN or from Graph-RISE. Their evidence: captioning CIDEr improves from 0.944 to 0.984 on Conceptual Captions test, and VQA accuracy improves from 51.9% to 53.7% on VizWiz test-standard, with the biggest per-type gains on number and other answers. They conclude that large-scale image-level semantic labels, even without scene-dissecting box annotations, improve current vision-and-language models when the evaluation images come from diverse domains.

Load-bearing premise

The load-bearing premise is that Graph-RISE's pretraining data and the Conceptual Captions and VizWiz evaluation images do not overlap enough to explain the gains; the paper never reports an overlap analysis, so the improvement could partly reflect data exposure rather than ultrafine-grained label quality.

Editorial extensions

If this is right

  • On Conceptual Captions test, replacing Faster R-CNN box features with Graph-RISE embeddings raises CIDEr from 0.944 to 0.984 and SPICE from 0.190 to 0.195.
  • On VizWiz test-standard, the same replacement raises VQA accuracy from 51.9% to 53.7%, including gains of +4.5 points on number questions and +3.3 points on other questions.
  • In captioning, global features, box features, and label embeddings are complementary: the full combination outperforms each single feature type.
  • Since box proposals are unchanged across the comparison, the gains isolate the featurization step rather than region selection.
  • Large-scale image-level semantic labels, which are cheaper than full box-and-label annotations, can improve out-of-domain captioning and VQA under this recipe.

Reading between the lines

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

  • One unresolved confound is the overlap between Graph-RISE's roughly 260 million web pretraining images and the web-sourced Conceptual Captions images; the paper does not measure it, so part of the gain could reflect pretraining-data exposure rather than ultrafine-grained label quality.
  • If decoupling proves robust, object-detection pretraining could be replaced by a shared web-scale featurizer plus any cheap class-agnostic proposer, removing the need to retrain detectors for each target domain.
  • A natural next experiment is the same featurizer swap on an in-domain benchmark such as COCO captioning or VQA2, where the Faster R-CNN detector is already well matched; the paper's transfer argument implies the gain will be smaller there.
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

3 major / 5 minor

Summary. The paper proposes decoupling object box proposal from region featurization in vision-and-language models. Instead of using Faster R-CNN features, the authors use the same Faster R-CNN proposals but featurize each cropped region with Graph-RISE, a graph-regularized image semantic embedding trained on roughly 260M web images with 40M noisy ultrafine-grained labels. They evaluate this on two transfer settings: image captioning on Conceptual Captions and visual question answering on VizWiz. The reported results show consistent gains: G+B-Ultra+L reaches 0.984 CIDEr versus 0.944 for G+B-FRCNN+L on the Conceptual Captions test set, and the Ultra-based VQA model scores 53.7% versus 51.9% on VizWiz test-standard. The paper argues that ultrafine-grained semantic labels improve the quality of bottom-up visual features, especially for out-of-domain images.

Significance. If the reported gains hold, the paper makes a useful contribution by showing that decoupling proposal and featurization allows large-scale weakly labeled data to improve downstream vision-and-language tasks, and by demonstrating an alternative to the standard Faster R-CNN features from Visual Genome. The strengths include evaluation on hidden test servers for both benchmarks, ablations isolating feature types, and qualitative examples that align with the claimed mechanism. However, the significance is tempered by the fact that only two benchmarks are used, all results are single runs without uncertainty estimates, and the featurizer is not released, which limits independent verification.

major comments (3)
  1. [Sections 3 and 4] The central transfer claim rests on an unexamined confound: Graph-RISE is trained on O(260M) web images (Section 3) and Conceptual Captions is composed of 3.3M web images with alt-text captions (Section 4), yet no overlap or near-duplicate analysis is reported between these sets. Because Graph-RISE features are not released, readers cannot independently check whether the 0.984 versus 0.944 test CIDEr gain for G+B-Ultra+L over G+B-FRCNN+L reflects genuine transfer from ultrafine-grained labels or simply the featurizer having seen the evaluation images or near-duplicates during pretraining. The authors should quantify overlap (e.g., via image hashing or embedding similarity on a dev sample) or otherwise rule out this explanation before the transfer interpretation is accepted.
  2. [Section 3 and Table 1] The comparison between B-FRCNN and B-Ultra conflates at least three factors: training-data scale (Visual Genome has roughly 108K images versus Graph-RISE's 260M images), label granularity (1,600 object and 400 attribute labels versus ultrafine-grained noisy labels), and feature dimensionality (2048D sparse versus 64D dense with additional projection layers). The paper's conclusion that ultrafine-grained semantic labels drive the improvement is not established unless the scale and architecture differences are controlled or explicitly discussed as a confound. A concrete test would be to compare against a Graph-RISE variant trained on a similarly small dataset, or to compare Faster R-CNN features trained on a larger weakly labeled set, so that the label-granularity hypothesis can be separated from the data-scale hypothesis.
  3. [Tables 1-3] All quantitative results are reported from single runs without error bars, confidence intervals, or significance tests. The key VizWiz aggregate difference is 1.8%, and the per-answer-type differences vary considerably (e.g., number +4.5% versus yes/no +1.4% and unanswerable +3.5% on test-standard), so without repeated runs or bootstrap estimates it is possible that part of the claimed improvement is within run-to-run variance. The authors should provide uncertainty estimates, at least for the dev-set numbers, and ideally for the test-set submissions if the evaluation server permits multiple submissions.
minor comments (5)
  1. [Section 3] There is a typo: 'More concretly' should be 'More concretely'.
  2. [Reproducibility] The paper does not provide a link to code, trained features, or the Graph-RISE model. Given that the central comparison depends on the Graph-RISE featurizer, releasing the extracted features for Conceptual Captions and VizWiz would substantially improve reproducibility and allow the overlap concern to be checked by the community.
  3. [Section 4 and Table 1] The test-set comparison includes only two submitted models (G+B-FRCNN+L and G+B-Ultra+L), while all ablations are reported on dev only. The paper states this, but it should be highlighted more explicitly that the headline comparison rests on a single pair of test submissions and that the other feature combinations are not compared on the hidden test set.
  4. [Figures 1 and 2] The qualitative examples are illustrative but may not be representative; a small quantitative error analysis (e.g., which answer categories or caption tokens improve most) would strengthen the claim that ultrafine-grained features specifically help with rare or long-tail concepts.
  5. [Abstract and Conclusion] The abstract and conclusion state that ultrafine-grained semantic labels lead to improved transfer learning, but the experiments only cover two benchmarks. Consider tempering the generalization claim or adding a sentence noting the narrowness of the empirical scope.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central captioning and VQA gains are measured on external challenge servers against fixed baselines, so the result does not reduce to the method's own fitted values.

full rationale

The paper's central claim is an empirical comparison of two featurization choices, Faster R-CNN box features versus Graph-RISE 'Ultra' box features, within otherwise fixed captioning and VQA models. The headline numbers are obtained by submitting to the Conceptual Captions Challenge server and the VizWiz Challenge server, i.e., external evaluation servers whose scores are not quantities fitted by the paper. Hyperparameters are tuned on dev splits, but the reported test results are not fitted values renamed as predictions. The Graph-RISE featurizer is taken from a prior preprint rather than derived in this paper, and the paper does not invoke a uniqueness theorem or define its inputs in terms of its outputs. The self-citations that exist, such as Sharma et al. 2018 for Conceptual Captions and the captioning model skeleton, concern the benchmark dataset and model architecture rather than the comparative conclusion; the benchmark itself is externally hosted, so these citations are not load-bearing for the transfer claim. A possible overlap between Graph-RISE's large web-image pretraining set and the web-sourced Conceptual Captions benchmark is a validity confound, not a definitional circularity, and it does not make the reported comparison equivalent to its inputs by construction. Therefore no circularity is identified.

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

No new mechanisms are invented; the paper depends on transfer properties of an existing embedding model and on benchmark validity. The main hidden premise is that pretraining data do not overlap with evaluation images.

free parameters (4)
  • Graph-RISE embedding dimension = 64
    Chosen in Graph-RISE, not by this paper; the compact 64D representation is central to the claimed featurization advantage.
  • FRCNN feature projection dimension in captioning = 64
    The authors report that projecting the 2048D FRCNN features to 64D gave the best captioning results, making the comparison dimension-matched but adding a tuned projection (Supplementary B).
  • Number of proposed boxes K = 100
    Fixed following Anderson et al.; the paper does not sweep it.
  • Box score threshold for captioning inputs = 0.001
    Regions with lower Faster R-CNN scores are removed before feature sequencing; the threshold is chosen without sensitivity analysis (Section 4).
assumptions (4)
  • domain assumption Faster R-CNN region proposals are of sufficient quality that decoupling leaves proposal recall unchanged.
    The method fixes the proposal module and only swaps the featurizer; if proposals were the limiting factor, the conclusion about featurization would not follow. See Section 3.
  • domain assumption Graph-RISE embeddings, trained on whole-image semantic labels, stay informative when applied to 224x224 region crops.
    B-Ultra features are produced by cropping and resizing proposed boxes before running Graph-RISE; the paper does not validate this crop-transfer behavior beyond the reported benchmarks. See Sections 3 and 4.
  • domain assumption The evaluation images in Conceptual Captions and VizWiz were not seen during Graph-RISE pretraining.
    Graph-RISE is trained on 260M web images and CC is also web-sourced; no overlap analysis is provided, so the transfer gains could be partly due to pretraining exposure. See Section 3 versus Section 4.
  • domain assumption The automatic captioning metrics used (CIDEr, ROUGE-L, SPICE) are accepted proxies for caption quality.
    Standard practice in the field; not justified or tested in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Decoupled Box Proposal and Featurization with Ultrafine-Grained Semantic Labels Improve Image Captioning and Visual Question Answering." pith.science (2026). https://pith.science/paper/DIJL72J3

@misc{pith2026190902097,
  author       = {Pith},
  title        = {Pith review of: Decoupled Box Proposal and Featurization with Ultrafine-Grained Semantic Labels Improve Image Captioning and Visual Question Answering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DIJL72J3}},
  note         = {Machine review of arXiv:1909.02097}
}
read the original abstract

Object detection plays an important role in current solutions to vision and language tasks like image captioning and visual question answering. However, popular models like Faster R-CNN rely on a costly process of annotating ground-truths for both the bounding boxes and their corresponding semantic labels, making it less amenable as a primitive task for transfer learning. In this paper, we examine the effect of decoupling box proposal and featurization for down-stream tasks. The key insight is that this allows us to leverage a large amount of labeled annotations that were previously unavailable for standard object detection benchmarks. Empirically, we demonstrate that this leads to effective transfer learning and improved image captioning and visual question answering models, as measured on publicly available benchmarks.

Figures

Figures reproduced from arXiv: 1909.02097 by the authors.

Figure 1
Figure 1. Ultrafine-grained semantic labels (at “in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Qualitative results from our image captioning models using [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Pipeline for converting an image to a sequence of image features in our highest performing image [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 42 canonical work pages

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Abadi, A

    M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Isard, Y. Jia, R. Jozefowicz, L. Kaiser, M. Kudlur, J. Levenberg, D. Man\' e , R. Monga, S. Moore, D. Murray, C. Olah, M. Schuster, J. Shlens, B. Steiner, I. Sutskever, K. Talwar, P. Tucker, V. Vanhouck...

    work page 2015

  4. [4]

    Peter Anderson, Basura Fernando, Mark Johnson, and Stephen Gould. 2016. SPICE: semantic propositional image caption evaluation. In ECCV

    work page 2016

  5. [5]

    Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney, Mark Johnson, Stephen Gould, and Lei Zhang. 2018. Bottom-up and top-down attention for image captioning and visual question answering. In Proceedings of CVPR

    work page 2018

  6. [6]

    Lawrence Zitnick, and Devi Parikh

    Stanislaw Antol, Aishwarya Agrawal, Jiasen Lu, Margaret Mitchell, Dhruv Batra, C. Lawrence Zitnick, and Devi Parikh. 2015. VQA : Visual question answering. In Proceedings of ICCV

    work page 2015

  7. [7]

    Yuhua Chen, Wen Li, Christos Sakaridis, Dengxin Dai, and Luc Van Gool. 2018. Domain adaptive Faster R-CNN for object detection in the wild. In Proceedings of CVPR

    work page 2018

  8. [8]

    Kun Fu, Junqi Jin, Runpeng Cui, Fei Sha, and Changshui Zhang. 2017. Aligning where to see and what to tell: Image captioning with region-based attention and scene-specific contexts. TPAMI, 39(12):2321--2334

    work page 2017

Show all 47 references
  1. [9]

    Golnaz Ghiasi, Tsung-Yi Lin, and Quoc V. Le. 2018. DropBlock : A regularization method for convolutional networks. In Proceedings of NeurIPS

    2018

  2. [10]

    Yash Goyal, Tejas Khot, Douglas Summers-Stay, Dhruv Batra, and Devi Parikh. 2017. Making the V in VQA matter: Elevating the role of image understanding in visual question answering. In Proceedings of CVPR

    2017

  3. [11]

    Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P

    Danna Gurari, Qing Li, Abigale J. Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P. Bigham. 2018. VizWiz Grand Challenge : Answering visual questions from blind people. In Proceedings of CVPR

    2018

  4. [12]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of CVPR

    2016

  5. [13]

    Judy Hoffman, Deepak Pathak, Eric Tzeng, Jonathan Long, Sergio Guadarrama, Trevor Darrell, and Kate Saenko. 2016. Large scale visual recognition through adaptation using joint representation and multiple instance learning. JMLR, 17(1):4954--4984

    2016

  6. [14]

    Hudson and Christopher D

    Drew A. Hudson and Christopher D. Manning. 2019. GQA : a new dataset for compositional question answering over real-world images. In Proceedings of CVPR

    2019

  7. [15]

    Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In Proceedings of ICML

    2015

  8. [16]

    Yu Jiang, Vivek Natarajan, Xinlei Chen, Marcus Rohrbach, Dhruv Batra, and Devi Parikh. 2018. Pythia v0.1: the winning entry to the VQA Challenge 2018. arXiv preprint arXiv:1807.09956

    2018 arXiv

  9. [17]

    Da-Cheng Juan, Chun-Ta Lu, Zhen Li, Futang Peng, Aleksei Timofeev, Yi-Ting Chen, Yaxi Gao, Tom Duerig, Andrew Tomkins, and Sujith Ravi. 2019. Graph-RISE : Graph-regularized image semantic embedding. arXiv preprint arXiv:1902.10814

    2019 arXiv

  10. [18]

    Jin-Hwa Kim, Yongseok Choi, Sungeun Hong, Jaehyun Jun, and Byoung-Tak Zhang. 2018. Bilinear attention networks for VizWiz challenge. In Proceedings of the ECCV Workshop on VizWiz Grand challenge

    2018

  11. [19]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In Proceedings of ICLR

    2015

  12. [20]

    Shamma, Michael Bernstein, and Li Fei-Fei

    Ranjay Krishna, Yuke Zhu, Oliver Groth, Justin Johnson, Kenji Hata, Joshua Kravitz, Stephanie Chen, Yannis Kalantidis, Li-Jia Li, David A. Shamma, Michael Bernstein, and Li Fei-Fei. 2017. Visual Genome : Connecting language and vision using crowdsourced dense image annotations...

    2017

  13. [21]

    Chin-Yew Lin and Franz Josef Och. 2004. Automatic evaluation of machine translation quality using longest common subsequence and skip-bigram statistics. In Proceedings of ACL

    2004

  14. [22]

    Belongie, Lubomir D

    Tsung - Yi Lin, Michael Maire, Serge J. Belongie, Lubomir D. Bourdev, Ross B. Girshick, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll \' a r, and C. Lawrence Zitnick. 2014. Microsoft COCO: common objects in context. In Proceedings of ECCV

    2014

  15. [23]

    Li Liu, Wanli Ouyang, Xiaogang Wang, Paul Fieguth, Jie Chen, Xinwang Liu, and Matti Pietik \"a inen. 2018. Deep learning for generic object detection: A survey. arXiv preprint arXiv:1809.02165

    2018 arXiv

  16. [24]

    Jiasen Lu, Jianwei Yang, Dhruv Batra, and Devi Parikh. 2018. Neural baby talk. In Proceedings of CVPR

    2018

  17. [25]

    Dhruv Mahajan, Ross Girshick, Vignesh Ramanathan, Kaiming He, Manohar Paluri, Yixuan Li, Ashwin Bharambe, and Laurens van der Maaten. 2018. Exploring the limits of weakly supervised pretraining. In Proceedings of ECCV

    2018

  18. [26]

    Corrado, and Jeff Dean

    Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S. Corrado, and Jeff Dean. 2013. Distributed representations of words and phrases and their compositionality. In Proceedings of NeurIPS

    2013

  19. [27]

    Miller, Richard Beckwith, Christiane Fellbaum, Derek Gross, and Katherine Miller

    George A. Miller, Richard Beckwith, Christiane Fellbaum, Derek Gross, and Katherine Miller. 1990. Introduction to WordNet : An on-line lexical database. International Journal of Lexicography, 3(4):235--244

    1990

  20. [28]

    Hyeonwoo Noh, Taehoon Kim, Jonghwan Mun, and Bohyung Han. 2019. Transfer learning via unsupervised task discovery for visual question answering. In Proceedings of CVPR

    2019

  21. [29]

    Marco Pedersoli, Thomas Lucas, Cordelia Schmid, and Jakob Verbeek. 2017. Areas of attention for image captioning. In Proceedings of ICCV

    2017

  22. [30]

    Gao Peng, Hongsheng Li, Haoxuan You, Zhengkai Jiang, Pan Lu, Steven Hoi, and Xiaogang Wang. 2019. Dynamic fusion with intra-and inter-modality attention flow for visual question answering. In Proceedings of CVPR

    2019

  23. [31]

    Joseph Redmon and Ali Farhadi. 2017. YOLO9000 : Better, faster, stronger. In Proceedings of CVPR

    2017

  24. [32]

    Mengye Ren, Ryan Kiros, and Richard Zemel. 2015 a . Exploring models and data for image question answering. In Proceedings of NeurIPS

    2015

  25. [33]

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. 2015 b . Faster R-CNN : Towards real-time object detection with region proposal networks. In Proceedings of NeurIPS

    2015

  26. [34]

    Berg, and Li Fei-Fei

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. 2015. ImageNet large scale visual recognition challenge. IJCV, 115(3):211--252

    2015

  27. [35]

    Tim Salimans and Durk P. Kingma. 2016. Weight normalization: A simple reparameterization to accelerate training of deep neural networks. In Proceedings of NeurIPS

    2016

  28. [36]

    Barrett, Mateusz Malinowski, Razvan Pascanu, Peter Battaglia, and Timothy Lillicrap

    Adam Santoro, David Raposo, David G. Barrett, Mateusz Malinowski, Razvan Pascanu, Peter Battaglia, and Timothy Lillicrap. 2017. A simple neural network module for relational reasoning. In Proceedings of NeurIPS

    2017

  29. [37]

    Piyush Sharma, Nan Ding, Sebastian Goodman, and Radu Soricut. 2018. Conceptual Captions : A cleaned, hypernymed, image alt-text dataset for automatic image captioning. In Proceedings of ACL

    2018

  30. [38]

    Chen Sun, Abhinav Shrivastava, Saurabh Singh, and Abhinav Gupta. 2017. Revisiting unreasonable effectiveness of data in deep learning era. In Proceedings of ICCV

    2017

  31. [39]

    Yuxing Tang, Josiah Wang, Xiaofang Wang, Boyang Gao, Emmanuel Dellandr \'e a, Robert Gaizauskas, and Liming Chen. 2017. Visual and semantic knowledge transfer for large scale semi-supervised object detection. TPAMI, 40(12):3045--3058

    2017

  32. [40]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Proceedings of NeurIPS

    2017

  33. [41]

    Lawrence Zitnick, and Devi Parikh

    Ramakrishna Vedantam, C. Lawrence Zitnick, and Devi Parikh. 2015. CIDEr : Consensus-based image description evaluation. In Proceedings of CVPR

    2015

  34. [42]

    Subhashini Venugopalan, Lisa Anne Hendricks, Marcus Rohrbach, Raymond Mooney, Trevor Darrell, and Kate Saenko. 2017. Captioning images with diverse objects. In Proceedings of CVPR

    2017

  35. [43]

    Baoyuan Wu, Weidong Chen, Yanbo Fan, Yong Zhang, Jinlong Hou, Junzhou Huang, Wei Liu, and Tong Zhang. 2019. Tencent ML-Images : A large-scale multi-label image database for visual representation learning. arXiv preprint arXiv:1901.01703

    2019 arXiv

  36. [44]

    Kelvin Xu, Jimmy Ba, Ryan Kiros, Kyunghyun Cho, Aaron Courville, Ruslan Salakhutdinov, Richard Zemel, and Yoshua Bengio. 2015. Show, attend and tell: Neural image caption generation with visual attention. In Proceedings of ICML

    2015

  37. [45]

    Yan Zhang, Jonathon Hare, and Adam Pr \"u gel-Bennett. 2018. Learning to count objects in natural images for visual question answering. In Proceedings of ICLR

    2018

  38. [46]

    a henb \

    Xingyi Zhou, Jiacheng Zhuo, and Philipp Kr \"a henb \"u hl. 2019. Bottom-up object detection by grouping extreme and center points. In Proceedings of CVPR

    2019

  39. [47]

    Yuke Zhu, Oliver Groth, Michael Bernstein, and Li Fei-Fei. 2016. Visual7W : Grounded question answering in images. In Proceedings of CVPR

    2016

Pith tools

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