Training-Time-Friendly Network for Real-Time Object Detection

Deng Cai; Guodong Xu; Haifeng Liu; Tu Zheng; Zheng Yang; Zili Liu

REVIEW 3 major objections 4 minor 30 references

Training-Time-Friendly Network for Real-Time Object Detection

T0 review · 3 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 The paper claims that a Gaussian-kernel encoding of dense training samples lets a real-time object detector train seven times faster than CenterNet without losing accuracy.

desk verdict A genuinely faster real-time detector with a solid ablation, but the batch-size analogy that motivates the speed-up is a heuristic, not a proven mechanism. read the letter →

arxiv 1909.00700 v3 pith:CXLO44IR submitted 2019-09-02 cs.CV cs.LGeess.IV

Zili Liu , Tu Zheng , Guodong Xu , Zheng Yang , Haifeng Liu , Deng Cai This is my paper · ORCID

classification cs.CVcs.LGeess.IV

keywords objectdetectionreal-timetrainingtimeanchor-freedetectorGaussiankernellearningratescalingMSCOCOsampleencoding

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 sets out to show that an object detector can be fast at inference, accurate, and quick to train all at once. It proposes TTFNet, a single-stage anchor-free detector that encodes a dense sub-area of regression targets around every annotated box with a Gaussian kernel, arguing this is analogous to enlarging the mini-batch size and therefore allows a larger learning rate and faster convergence. On MS COCO the detector is reported to reduce training time by more than seven times compared with CenterNet and other real-time detectors while keeping or improving accuracy. The implication is that high-accuracy real-time detection no longer requires a long training schedule or heavy data augmentation.

What carries the argument

The load-bearing mechanism is a reinterpretation of the SGD update in terms of individual training samples: $w_{t+1}=w_t-\eta\frac{1}{nm}\sum_{s\in B}\nabla l(s,w_t)$, combined with the empirical linear scaling rule for learning rates. The paper argues that generating more samples per annotated box is similar to increasing the batch size, so a learning-rate multiplier is justified. The Gaussian kernel $K(x,y)=\exp\!\left(-\frac{(x-x_0)^2}{2\sigma_x^2}-\frac{(y-y_0)^2}{2\sigma_y^2}\right)$ defines both the localization activation and the regression sample area, and the per-sample weight $W_{ij}=\log(a_m)\,G_m(i,j)/\sum_{(x,y)\in A_m}G_m(x,y)$ balances object scales while emphasizing center-near pixels. Regression is supervised by GIoU loss on decoded boxes.

What would settle it

Train the same network with and without the learning-rate increase while keeping the Gaussian sample encoding fixed: if convergence speed is identical, the claimed batch-size analogy is not responsible for the speedup.

Watch

Extended reading notes

Core claim

The central claim is that the number of supervised samples encoded from each annotated box plays the same role as the mini-batch size in stochastic gradient descent, so multiplying the sample count justifies multiplying the learning rate and shortens the training schedule. Concretely, the paper's TTFNet uses a Gaussian kernel centered on each box to define a sub-area in which every pixel predicts the distance to the four box sides, turning one regression target into many. Sample weights combine the logarithm of box area with normalized Gaussian probability, preserving information from large boxes without drowning out small ones, and an aspect-ratio-aware Gaussian is used for center localization. The reported results include 35.1 AP for TTFNet-53 after 6.1 hours of training and 32.9 AP for a super-fast version after 3.1 hours, compared with CenterNet's 140-epoch schedule.

Load-bearing premise

The paper assumes that adding many correlated training samples from the same annotated box is equivalent, for the purpose of setting the learning rate, to adding independent images to the mini-batch.

Editorial extensions

If this is right

If the sample-count/batch-size analogy holds, any anchor-free detector can shorten its training schedule by densely encoding annotated boxes, not just TTFNet.
Training-time-sensitive applications such as neural architecture search become feasible for real-time detection, since candidate networks can be evaluated in hours rather than days.
From-scratch training on COCO reaches competitive accuracy (30.3 AP for ResNet-18) without ImageNet pretraining, which matters when pretrained weights are unavailable.
Detectors equipped with this encoding can drop heavy augmentation and long schedules, making experiments reproducible with modest GPU budgets.

Reading between the lines

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

The linear-scaling analogy is only argued qualitatively ($1\le l\le k$); a sharper statement would be to measure the effective gradient noise variance as sample count grows and predict when the analogy breaks because of intra-image correlation.
The Gaussian sub-area is a form of soft label assignment for regression; the same idea could transfer to other dense prediction tasks such as keypoint estimation or depth prediction, where center-anchored targets may be under-sampled.
One testable extension: instead of enlarging learning rate, keep it fixed and reduce the number of iterations; if the sample-count effect is genuine, convergence curves should collapse onto a single curve when plotted against total samples seen.

Share X Bluesky LinkedIn Reddit HN

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

First, the honest takeaway: this is an empirically strong paper. TTFNet delivers a real-time detector that trains about seven times faster than CenterNet and much faster than SSD/YOLOv3, with accuracy that is competitive or better. The core new idea — Gaussian-kernel encoding of dense regression samples, with sample weights based on log-area and Gaussian probability — is genuinely different from CenterNet's single-pixel regression, FCOS's whole-box, and FoveaBox's rectangle. The ablations are careful: they show the contribution of the Gaussian kernel, the aspect ratio, the weighting scheme, and the shortcut connections. The code is public. This is a solid practical contribution.

The weak point is the stated mechanism. The paper argues that encoding more samples per annotated box is like increasing the batch size, so the learning rate can be enlarged with a factor l between 1 and k. That is an unproven extrapolation of the linear scaling rule from independent images to strongly correlated samples inside one image. Equation (3) just rewrites the SGD update; it does not establish that the gradient noise behaves like that of a larger mini-batch. The authors do hedge — "qualitatively draw a similar conclusion" — so the claim is a heuristic, not a theorem. Table 5, which is supposed to validate it, is confounded: β controls both the number of regression samples and the shape of the target distribution, and the learning rate is varied jointly with β. No condition fixes the total number of samples while varying the number of images. Moreover, for β=0.2 at the 1x schedule, AP drops from 31.0 at LR=6e-3 to 30.6 at LR=1.8e-2, which contradicts the idea that more samples always permit a larger learning rate. So the empirical speed-up is real, but the explanation given for it is not established.

That said, the paper's contribution does not rest on the mechanism being true. The fast training could come from the dense supervision, the Gaussian weighting, the reduction of ambiguous samples, or the shortcut connections — all ablated. The reported training-time numbers are reasonable; they are not fully specified but are on the same hardware and benchmark, and Table 8's direct CenterNet comparison is informative.

This paper deserves a serious peer review. The technique is novel, the evaluation is meaningful, and the limitations are honestly flagged. I would recommend accepting with a request to reframe the batch-size analogy as a heuristic rather than a proven equivalence, and ideally a cleaner experiment that holds the total number of regression samples constant. For researchers in real-time detection or training-time-sensitive applications, this is a useful and citable work.

Referee Report

3 major / 4 minor

Summary. The paper proposes TTFNet, an anchor-free, single-stage, light-head object detector designed to balance training time, inference speed, and accuracy. The core idea is to encode multiple dense training samples from each annotated box using Gaussian kernels for both localization and size regression, and to argue that this is analogous to increasing the mini-batch size, thereby allowing a larger learning rate and faster convergence. The authors present ablations on sample weighting, Gaussian kernel size, aspect-ratio-aware kernels, and shortcut connections, and report experiments on MS COCO showing that TTFNet reduces training time by more than seven times compared with CenterNet and other real-time detectors while retaining competitive accuracy, with versions based on ResNet-18/34 and DarkNet-53.

Significance. If the empirical results are correct, the paper delivers a practically valuable result: a real-time detector that reaches strong MS COCO accuracy with dramatically reduced training time. The release of code, the use of an external benchmark, and the extensive ablation study are clear strengths. The central mechanism—that encoding more correlated samples per annotated box behaves like enlarging the mini-batch size for learning-rate scaling—is only argued qualitatively and is not rigorously established by the experiments as presented. Nevertheless, the practical contribution of the detector design and its training-time/accuracy tradeoff appear to stand on their own, and the mechanism question, while important, is a correctness-risk concern rather than a demonstrated flaw in the empirical claims.

major comments (3)

[Motivation] The paper assumes without proof that multiplying the number of encoded samples per image by k justifies multiplying the learning rate by l with 1 <= l <= k, extending Goyal et al.'s linear scaling rule from independent images to strongly correlated samples within one image. Equation (3) is only a rewriting of the SGD update; it does not establish that the gradient noise behaves as if nm independent samples were drawn. Since the learning-rate enlargement is the stated mechanism for faster training, this is a load-bearing gap. Please either provide a derivation or argument under explicit assumptions, or design an experiment that isolates the effect of sample count from image count (e.g., fixing total regression samples while varying the number of images, or measuring gradient-noise variance as a function of beta).
[Ablation Study] The validation of the learning-rate scaling claim is confounded. The table varies beta and learning rate simultaneously, and the 1x/2x schedules change the number of epochs; there is no condition that holds the total number of regression samples fixed while varying the number of images, nor one that varies sample count while keeping the number of images constant. Moreover, for beta=0.2 at the 1x schedule, AP drops from 31.8 at LR=1.2e-2 to 30.6 at LR=1.8e-2, which is inconsistent with the statement that larger beta guarantees a larger usable learning rate. A clean factorial design (or explicit control of the effective sample count) is needed to support the claimed equivalence.
[Motivation and Ablation Study] The paper claims that CenterNet's slow convergence is mainly caused by its single regression sample per box, but this attribution is not directly tested. Figure 1 changes the learning rate and removes data augmentation, but does not vary the number of regression samples, and Table 8 compares different network architectures with different upsampling and head designs. An experiment that keeps the network architecture fixed and varies only the regression sample encoding (e.g., a CenterNet-style head with multiple Gaussian-encoded samples) is needed to support the causal claim that the sample-count increase is the decisive factor.

minor comments (4)

[Ablation Study] The text refers to 'super-fast TTNet-53' in the first paragraph; this should be 'TTFNet-53'.
[Overall Design] The phrase 'connnection' in the shortcut-connection paragraph is a typo and should be 'connection'.
[Gaussian Kernels for Training] The symbols alpha_f and beta_f are used in the modified focal loss but are not defined in the surrounding text; please define them explicitly.
[Ablation Study] The column headers of Table 1 (Norm, Sqrt, Log, Gaussian) with checkmarks do not make clear which configuration corresponds to each reported AP; please clarify the mapping.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: TTFNet's claims rest on external COCO evaluation and standard model selection; the sample-count/batch-size analogy is an unproven assumption but not a self-referential derivation.

full rationale

The paper's central claim is an empirical detector evaluated on the external MS COCO val split, with training-time comparisons against CenterNet, SSD, and YOLOv3; no predicted quantity is defined in terms of a fitted parameter. The motivation argues qualitatively, via Eqs. (1)-(3) and Goyal et al.'s linear scaling rule, that encoding more regression samples is 'similar' to increasing batch size, explicitly hedging with 'we can qualitatively draw a similar conclusion'; Eq. (3) is only an algebraic rewriting of the SGD update and does not construct the conclusion from its own target. This is an unverified modeling assumption (a correctness risk if the equivalence fails), not circularity. Hyperparameters α, β, and sample weights are chosen by ablations on COCO val, which is conventional model selection rather than fitting the answer into the input. No load-bearing self-citation chain is present: the cited CenterNet, Goyal et al., and mmdetection results are external, and the detector's AP/FPS/training-time numbers are independently measured on the benchmark. Even if Table 5 confounds β and learning rate, that weakness concerns evidence quality, not a reduction of output to input, so no circular step is exhibited.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim depends on several selected hyperparameters (alpha, beta, s, loss weights, shortcut depths) and on two domain assumptions about learning-rate scaling. The paper does not introduce new physical or conceptual entities. The weakest link is the extension of the linear scaling rule to correlated samples within one image.

free parameters (5)

alpha (localization Gaussian size) = 0.54
Controls the standard deviation of the localization heatmap Gaussian (sigma_x = alpha*w/6, sigma_y = alpha*h/6). The paper states it is not carefully selected.
beta (regression Gaussian size) = 0.54
Controls the size of the Gaussian sub-area used for regression samples and is set equal to alpha so the same kernel can be shared. It is effectively tuned via Tables 2-5.
s (prediction scaling) = 16
A fixed scalar that enlarges predicted box sizes for easier optimization, as described after Eq. (5).
loss weights wloc, wreg = 1.0 and 5.0
Manually chosen weights for localization and regression losses, set in the Total Loss paragraph.
shortcut connection depth (3,2,1) = 3, 2, 1 for stages 2,3,4
Chosen to balance speed and accuracy; the paper says it is not carefully selected.

assumptions (3)

domain assumption Linear scaling rule for SGD with large mini-batches (Goyal et al. 2017) holds for the training-sample analog.
The Motivation section uses Eqs. (1)-(3) to equate more encoded samples with a larger mini-batch, assuming gradients change slowly across the iterations being merged. The paper admits the conclusion is qualitative.
domain assumption Samples encoded from the same image, despite correlation, provide information with differences sufficient to behave like independent samples for learning-rate scaling.
Stated in the Motivation section after Eq. (3): 'Although the encoded samples have a strong correlation, they are still able to contribute information with differences.'
ad hoc to paper CenterNet's slow convergence is primarily caused by its single regression sample per box.
The experiment in Figure 1 removes augmentation and raises the learning rate but does not directly test this attribution. The conclusion is used to justify the Gaussian sub-area design and is therefore specific to this paper's narrative.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Training-Time-Friendly Network for Real-Time Object Detection." pith.science (2026). https://pith.science/paper/CXLO44IR

@misc{pith2026190900700,
  author       = {Pith},
  title        = {Pith review of: Training-Time-Friendly Network for Real-Time Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CXLO44IR}},
  note         = {Machine review of arXiv:1909.00700}
}

read the original abstract

Modern object detectors can rarely achieve short training time, fast inference speed, and high accuracy at the same time. To strike a balance among them, we propose the Training-Time-Friendly Network (TTFNet). In this work, we start with light-head, single-stage, and anchor-free designs, which enable fast inference speed. Then, we focus on shortening training time. We notice that encoding more training samples from annotated boxes plays a similar role as increasing batch size, which helps enlarge the learning rate and accelerate the training process. To this end, we introduce a novel approach using Gaussian kernels to encode training samples. Besides, we design the initiative sample weights for better information utilization. Experiments on MS COCO show that our TTFNet has great advantages in balancing training time, inference speed, and accuracy. It has reduced training time by more than seven times compared to previous real-time detectors while maintaining state-of-the-art performances. In addition, our super-fast version of TTFNet-18 and TTFNet-53 can outperform SSD300 and YOLOv3 by less than one-tenth of their training time, respectively. The code has been made available at \url{https://github.com/ZJULearning/ttfnet}.

Figures

Figures reproduced from arXiv: 1909.00700 by the authors.

**Figure 3.** Different strategies for defining training samples. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 13 canonical work pages

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...
[2]

C.; and Lin, D

Chen, K.; Wang, J.; Pang, J.; Cao, Y.; Xiong, Y.; Li, X.; Sun, S.; Feng, W.; Liu, Z.; Xu, J.; Zhang, Z.; Cheng, D.; Zhu, C.; Cheng, T.; Zhao, Q.; Li, B.; Lu, X.; Zhu, R.; Wu, Y.; Dai, J.; Wang, J.; Shi, J.; Ouyang, W.; Loy, C. C.; and Lin, D. 2019. Mmdetection: Open mmlab detection toolbox and benchmark. CoRR abs/1906.07155

arXiv 2019
[3]

Deng, J.; Dong, W.; Socher, R.; Li, L.-J.; Li, K.; and Fei-Fei, L. 2009. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition , 248--255. Ieee

2009
[4]

Duan, K.; Bai, S.; Xie, L.; Qi, H.; Huang, Q.; and Tian, Q. 2019. Centernet: Keypoint triplets for object detection. CoRR abs/1904.08189

arXiv 2019
[5]

Gao, Y.; Ma, J.; Zhao, M.; Liu, W.; and Yuille, A. L. 2019. NDDR - CNN : Layerwise feature fusing in multi-task cnns by neural discriminative dimensionality reduction. In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR)

work page 2019
[6]

Ghiasi, G.; Lin, T.-Y.; and Le, Q. V. 2019. Nas-fpn: Learning scalable feature pyramid architecture for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 7036--7045

work page 2019
[7]

B.; Noordhuis, P.; Wesolowski, L.; Kyrola, A.; Tulloch, A.; Jia, Y.; and He, K

Goyal, P.; Doll \' a r, P.; Girshick, R. B.; Noordhuis, P.; Wesolowski, L.; Kyrola, A.; Tulloch, A.; Jia, Y.; and He, K. 2017. Accurate, large minibatch SGD: training imagenet in 1 hour. CoRR abs/1706.02677

arXiv 2017
[8]

Huang, L.; Yang, Y.; Deng, Y.; and Yu, Y. 2015. Densebox: Unifying landmark localization with end to end object detection. CoRR abs/1509.04874

arXiv 2015

Show all 30 references

[9]

Ioffe, S., and Szegedy, C. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015 , 448--456

2015
[10]

Kong, T.; Sun, F.; Liu, H.; Jiang, Y.; and Shi, J. 2019. Foveabox: Beyond anchor-based object detector. CoRR abs/1904.03797

2019 arXiv
[11]

Law, H., and Deng, J. 2018. Cornernet: Detecting objects as paired keypoints. In Proceedings of the European Conference on Computer Vision (ECCV) , 734--750

2018
[12]

Lin, T.-Y.; Maire, M.; Belongie, S.; Hays, J.; Perona, P.; Ramanan, D.; Doll \'a r, P.; and Zitnick, C. L. 2014. Microsoft coco: Common objects in context. In European conference on computer vision , 740--755. Springer

2014
[13]

Lin, T.-Y.; Doll \'a r, P.; Girshick, R.; He, K.; Hariharan, B.; and Belongie, S. 2017a. Feature pyramid networks for object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition , 2117--2125
[14]

Lin, T.-Y.; Goyal, P.; Girshick, R.; He, K.; and Doll \'a r, P. 2017b. Focal loss for dense object detection. In Proceedings of the IEEE international conference on computer vision , 2980--2988
[15]

Liu, W.; Anguelov, D.; Erhan, D.; Szegedy, C.; Reed, S.; Fu, C.-Y.; and Berg, A. C. 2016. Ssd: Single shot multibox detector. In European conference on computer vision , 21--37. Springer

2016
[16]

Redmon, J., and Farhadi, A. 2017. Yolo9000: better, faster, stronger. In Proceedings of the IEEE conference on computer vision and pattern recognition , 7263--7271

2017
[17]

Redmon, J., and Farhadi, A. 2018. Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767

2018 arXiv
[18]

Redmon, J.; Divvala, S.; Girshick, R.; and Farhadi, A. 2016. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition , 779--788

2016
[19]

Ren, S.; He, K.; Girshick, R.; and Sun, J. 2015. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems , 91--99

2015
[20]

Rezatofighi, H.; Tsoi, N.; Gwak, J.; Sadeghian, A.; Reid, I.; and Savarese, S. 2019. Generalized intersection over union: A metric and a loss for bounding box regression. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 658--666

2019
[21]

Tian, Z.; Shen, C.; Chen, H.; and He, T. 2019. FCOS: fully convolutional one-stage object detection. CoRR abs/1904.01355

2019 arXiv
[22]

C.; and Lin, D

Wang, J.; Chen, K.; Yang, S.; Loy, C. C.; and Lin, D. 2019a. Region proposal by guided anchoring. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2965--2974
[23]

Wang, N.; Gao, Y.; Chen, H.; Wang, P.; Tian, Z.; and Shen, C. 2019b. NAS-FCOS: fast neural architecture search for object detection. CoRR abs/1906.04423

1906 arXiv
[24]

Yu, J.; Jiang, Y.; Wang, Z.; Cao, Z.; and Huang, T. 2016. Unitbox: An advanced object detection network. In Proceedings of the 24th ACM international conference on Multimedia , 516--520. ACM

2016
[25]

Yu, F.; Wang, D.; Shelhamer, E.; and Darrell, T. 2018. Deep layer aggregation. In 2018 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2018, Salt Lake City, UT, USA, June 18-22, 2018 , 2403--2412. IEEE Computer Society

2018
[26]

a henb \

Zhou, X.; Wang, D.; and Kr \" a henb \" u hl, P. 2019. Objects as points. CoRR abs/1904.07850

2019 arXiv
[27]

Zhu, X.; Hu, H.; Lin, S.; and Dai, J. 2019. Deformable convnets v2: More deformable, better results. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 9308--9316

2019
[28]

Zhu, C.; He, Y.; and Savvides, M. 2019. Feature selective anchor-free module for single-shot object detection. CoRR abs/1903.00621

2019 arXiv
[29]

Zoph, B., and Le, Q. V. 2017. Neural architecture search with reinforcement learning. In 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings

2017
[30]

Zoph, B.; Vasudevan, V.; Shlens, J.; and Le, Q. V. 2018. Learning transferable architectures for scalable image recognition. In 2018 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2018, Salt Lake City, UT, USA, June 18-22, 2018 , 8697--8710

2018

Pith tools

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

[1] [1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

[2] [2]

C.; and Lin, D

Chen, K.; Wang, J.; Pang, J.; Cao, Y.; Xiong, Y.; Li, X.; Sun, S.; Feng, W.; Liu, Z.; Xu, J.; Zhang, Z.; Cheng, D.; Zhu, C.; Cheng, T.; Zhao, Q.; Li, B.; Lu, X.; Zhu, R.; Wu, Y.; Dai, J.; Wang, J.; Shi, J.; Ouyang, W.; Loy, C. C.; and Lin, D. 2019. Mmdetection: Open mmlab detection toolbox and benchmark. CoRR abs/1906.07155

arXiv 2019

[3] [3]

Deng, J.; Dong, W.; Socher, R.; Li, L.-J.; Li, K.; and Fei-Fei, L. 2009. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition , 248--255. Ieee

2009

[4] [4]

Duan, K.; Bai, S.; Xie, L.; Qi, H.; Huang, Q.; and Tian, Q. 2019. Centernet: Keypoint triplets for object detection. CoRR abs/1904.08189

arXiv 2019

[5] [5]

Gao, Y.; Ma, J.; Zhao, M.; Liu, W.; and Yuille, A. L. 2019. NDDR - CNN : Layerwise feature fusing in multi-task cnns by neural discriminative dimensionality reduction. In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR)

work page 2019

[6] [6]

Ghiasi, G.; Lin, T.-Y.; and Le, Q. V. 2019. Nas-fpn: Learning scalable feature pyramid architecture for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 7036--7045

work page 2019

[7] [7]

B.; Noordhuis, P.; Wesolowski, L.; Kyrola, A.; Tulloch, A.; Jia, Y.; and He, K

Goyal, P.; Doll \' a r, P.; Girshick, R. B.; Noordhuis, P.; Wesolowski, L.; Kyrola, A.; Tulloch, A.; Jia, Y.; and He, K. 2017. Accurate, large minibatch SGD: training imagenet in 1 hour. CoRR abs/1706.02677

arXiv 2017

[8] [8]

Huang, L.; Yang, Y.; Deng, Y.; and Yu, Y. 2015. Densebox: Unifying landmark localization with end to end object detection. CoRR abs/1509.04874

arXiv 2015