Utilizing Temporal Information in Deep Convolutional Network for Efficient Soccer Ball Detection and Tracking

Anna Kukleva; Hafez Farazi; Mohammad Asif Khan; Sven Behnke

REVIEW 3 major objections 5 minor 26 references

Utilizing Temporal Information in Deep Convolutional Network for Efficient Soccer Ball Detection and Tracking

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 Soccer ball detection improves when the detector sees a 20-frame history: recall rises to 0.987 and F1 to 0.981 over the single-frame baseline, with a temporal convolution adding only 1.1 ms.

desk verdict The central claim that temporal context improves detection is confounded by extra real-data fine-tuning; the paper is solid engineering with a useful dataset but needs a matched baseline before the F1 gain means anything. read the letter →

arxiv 1909.02406 v2 pith:VHORBQQI submitted 2019-09-05 cs.CV

Anna Kukleva , Mohammad Asif Khan , Hafez Farazi , Sven Behnke This is my paper · ORCID

classification cs.CV

keywords robocupdeeplearningballdetectionfullyconvolutionalneuralnetworkspatio-temporaltemporalConvLSTMobjecttracking

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 aims to establish that a soccer-ball detector for a humanoid robot works better when it sees a short history of frames than when it sees only one frame. Its recipe is to take a fully convolutional single-image heatmap detector (SweatyNet), pipe the last 20 heatmaps into a temporal layer, and learn to predict the next heatmap. On its own test data the temporal models raise recall from 0.972 to 0.985–0.987 and F1 from 0.973 to 0.980–0.981, and the feed-forward TCN variant does this in only 1.1 ms of added inference time. If true, that gives RoboCup robots a real-time way to keep tracking a ball through occlusion and blur without paying the latency of a recurrent network.

What carries the argument

The load-bearing mechanism is an encoder-decoder heatmap network extended by a temporal layer and a learned residual gate. The base network (SweatyNet-1) produces a 160x120 probability map; meanwhile the outputs of its second and sixth encoder blocks are concatenated, passed through a 7x7 convolution, multiplied by a learnable weight $w$, and added element-wise to the base heatmap. This gated residual is the input to the temporal layer—a TCN made of dilated causal convolutions, or a ConvLSTM, or a ConvGRU—which consumes $h=20$ heatmaps and outputs the prediction for the next frame. The TCN's dilated causal convolutions are what make the practical claim credible: they extend the receptive field over time in a feed-forward manner, capturing long-range dependencies at a cost of only 1.1 ms.

What would settle it

Take a fresh RoboCup video sequence, hold out all of its frames from training and fine-tuning, freeze the base SweatyNet weights, and compare single-image detection against temporal detection on those frames; if temporal recall and F1 do not exceed SweatyNet's on the held-out sequence, the claimed benefit of history collapses.

Watch

Extended reading notes

Core claim

The central claim is that spatiotemporal context, not just the current image, determines where the ball is: given the network's own heatmaps for frames $t-h$ to $t-1$ ($h=20$), a temporal model predicts the heatmap at $t$ and does so more accurately than the same single-frame network. The decisive comparison is SweatyNet-1 with dropout 0.5 (recall 0.972, F1 0.973) against the same network with a temporal layer (best recall 0.987, best F1 0.981). The authors also claim that pre-training the temporal stack on artificially generated ball trajectories and then fine-tuning it on real RoboCup 2018 sequences improves over fine-tuning on real data alone, and that TCN matches or nearly matches ConvLSTM and ConvGRU accuracy while being far faster.

Load-bearing premise

The load-bearing premise is that the temporal models were evaluated on a held-out test set that does not overlap with the real RoboCup 2018 sequences used to fine-tune them; the paper does not state that temporal split.

Editorial extensions

If this is right

A robot can keep a ball estimate through short occlusions and missed detections, because the temporal model continues to produce a heatmap when the single-frame network's confidence is low.
For real-time RoboCup play, TCN is the preferred temporal choice: it gives the same detection quality as recurrent variants at a fraction of their added latency.
Adding synthetic pre-training to the temporal stack reduces the amount of labeled real video needed, improving results over real-only fine-tuning.
The trained network generalizes to two balls in one frame even though it was trained on single-ball sequences.
The same architecture could be reused to detect other soccer objects such as goalposts and robots, as the authors note in their conclusion.

Reading between the lines

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

Because the temporal layer consumes heatmaps rather than raw pixels, the same residual-gated design could in principle be attached to any single-frame heatmap detector; the paper only demonstrates it with SweatyNet.
An ablation that freezes the base SweatyNet weights while training only the temporal layer would separate the contribution of trajectory history from the contribution of extra fine-tuning on RoboCup footage; the paper does not report that split.
A natural next experiment is to vary the history length $h$ and prediction horizon $p$; the paper fixes $h=20, p=1$, so the minimum context needed to recover a ball after occlusion remains unknown.

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

Main take: the empirical case that temporal context improves soccer ball detection is not made. In Table 2(b), the temporal models are fine-tuned on real RoboCup 2018 sequences, while the SweatyNet-1 baseline is taken directly from Table 1 with no additional fine-tuning. That changes two variables at once, and the reported gain is only about 0.008 F1. The stress-test note is correct: you need a single-image SweatyNet-1 fine-tuned on the same real sequences with the same schedule, then compare. Without that control, the improvement could just be extra training data or domain adaptation, not temporal history.

What is genuinely useful: the SoccerData dataset, the public code, and the careful engineering for real-time use. TCN adding only 1.1 ms on the robot is a real result, and the authors honestly show that ConvLSTM and ConvGRU are far slower. The consistent direction across all three temporal models is suggestive, even if small. The qualitative result on detecting two balls after training with one is a nice generalization check.

Soft spots, in order of severity. First, the confounded baseline: load-bearing, and fixable with one experiment. Second, no error bars or significance tests, which matters because the effect is tiny. Third, the temporal train/test split is never stated, so we cannot rule out overlap between fine-tuning sequences and test frames. Fourth, the occlusion claim rests largely on synthetic data, with only one qualitative real example. These are not fatal; they are addressable.

This is honest engineering work. The dataset and speed numbers are useful to the RoboCup community, and the paper deserves a serious referee, but the headline claim about temporal context needs a matched baseline before it can be believed. If I were reviewing, I would request that experiment and some measure of variance, then the paper would be publishable as a useful systems-plus-dataset contribution.

Who for: RoboCup participants and people building small-robot vision systems. A general computer vision audience gets little, but the dataset and the TCN speed finding are worth having on record.

Referee Report

3 major / 5 minor

Summary. The paper proposes extending a single-image fully convolutional soccer-ball detector (SweatyNet) with temporal sequence models—TCN, ConvLSTM, and ConvGRU—to exploit the history of frames for ball detection and tracking. The authors introduce SoccerData, a new dataset of 4562 annotated images, and also extract real RoboCup 2018 sequences for fine-tuning. The temporal models are pre-trained on synthetic sequences and/or fine-tuned on real sequences, and the reported results show small F1 improvements over the SweatyNet-1 baseline (F1 0.973 to 0.980–0.981) together with an added TCN inference cost of only 1.1 ms. The central claim is that temporal context improves detection, especially in challenging and occluded scenarios.

Significance. If the central claim were established, the work would be a useful, low-cost enhancement for real-time soccer ball detection in RoboCup, and the released dataset and code would support reproducibility. The paper honestly reports three temporal architectures, distinguishes pretraining from direct fine-tuning, and measures inference time on the actual robot platform. The main weakness is that the quantitative evidence for the claim is weak and confounded: the comparison in Table 2(b) changes both the architecture and the training data at once, the training/test split for the real sequences is not documented, and no variance or significance information is provided. The reported effect is small, so these issues are load-bearing for the paper's main conclusion rather than cosmetic.

major comments (3)

[Section 4.1, Table 2(b)] The central comparison is confounded by training-data differences. The SweatyNet-1 (0.5) row in Table 2(b) is numerically identical to the SweatyNet-1 (0.5) test row in Table 1, which was trained only on SoccerData images. In contrast, Section 4.1 states that the temporal models are fine-tuned on real sequences extracted from RoboCup 2018 games, and Table 2(b) says 'real denotes that training of the sequential part is performed only on real data.' Thus the temporal rows benefit from both the temporal architecture and additional fine-tuning on a new real-data domain. The reported F1 gain of about 0.008 could therefore be due to extra data or domain adaptation rather than temporal context. A necessary control is a single-image SweatyNet-1 fine-tuned on the same real RoboCup sequences with the same schedule, without any temporal layer; that control is absent.
[Section 4.1, Section 5, Table 2(b)] The temporal train/test split is not stated. The paper does not say whether the 20 sets of consecutive real frames (average length 60) are split into training and test sets, how many sequences are used for fine-tuning versus evaluation, or whether any test frames overlap with fine-tuning frames. Without this information, the test-set metrics in Table 2(b) cannot be interpreted as held-out performance. The test split should be described explicitly, and a guarantee that no fine-tuning frames overlap with the test frames should be stated.
[Section 5, Table 2(b)] No variance, error bars, or significance tests are reported, despite the small effect size. The F1 differences are 0.973 versus 0.980–0.981, and all values come from a single split. Given typical run-to-run variation in deep-network training, these differences could easily be within noise. Reporting results over multiple random initializations or per-sequence evaluation would be needed to support the claim that the temporal models 'give an improvement over the SweatyNet.'

minor comments (5)

[Section 1] There is an orphan fragment 'research 1.' immediately after the sentence about releasing data and implementation on GitHub; this appears to be a leftover footnote marker and should be removed or converted into a proper footnote.
[Section 4.1] The word 'experimet' should be 'experiment' in the sentence 'In the second experimet, we evaluated our temporal approach.'
[Algorithm 1] The postprocessing algorithm uses thresholds M > 0.1 and a minimum contour area A_min, but the chosen numerical values for these parameters are not reported. These should be specified for reproducibility.
[Table 2(a)] The caption says 'For sequential models, we report time on top of the base model,' but it is unclear whether the reported times are additive overhead or total inference time. Please clarify the timing protocol.
[Figures 2, 4, and 5] The multiple-ball generalization is supported only by qualitative examples. A quantitative evaluation on the two-ball scenario, even a small one, would strengthen the generalization claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the paper's claims are empirical comparisons, not results derived from their own definitions.

full rationale

The paper contains no first-principles derivation whose output is equivalent to its input by construction. Its central claim, that adding TCN, ConvLSTM, or ConvGRU layers improves soccer-ball detection over a SweatyNet baseline, is supported by training and evaluation on data with reported metrics (FDR, PR, RC, F1, Acc) in Table 2(b). The temporal models use the authors' prior SweatyNet as a feature extractor and baseline; this self-citation is transparent and does not make the comparison circular. The SweatyNet baseline is an external prior model from Schnekenburger et al. (2017), while the temporal extensions are newly added architectures trained separately. No equation is defined in terms of the quantity it is claimed to predict, and no fitted parameter is renamed as a prediction. The one legitimate concern is experimental control: the temporal rows are fine-tuned on real RoboCup 2018 sequences, while the SweatyNet-1 (0.5) baseline row appears to be the model from Table 1 trained only on SoccerData, so the reported F1 gain may reflect additional training data rather than temporal context. That is a validity and confound concern, not circularity, because it does not reduce a predicted quantity to an input by construction. Therefore no circular step is identified.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical systems paper, so its load-bearing assumptions are data fidelity, transferability of synthetic training, and evaluation comparability rather than mathematical axioms. The free parameters are window sizes, thresholds, and the learned residual gate, all of which are either hand-chosen or fitted and directly influence the reported metrics.

free parameters (6)

history length h = 20
Hand-chosen window of past frames for all sequence experiments; the claimed benefit of temporal context depends on this length.
prediction horizon p = 1
Hand-chosen output length; the method predicts only one future heatmap, so the evaluation measures detection, not long-horizon tracking.
residual learnable weight w = 0.57 with pretraining, 0.49 without
Learned scalar that gates high-resolution information from early SweatyNet layers into the temporal model; reported average values in Section 5.
true-positive distance threshold gamma = 5
Hand-chosen pixel distance between predicted and ground-truth center that defines all precision, recall, and F1 metrics (Section 4.3).
minimum contour area A_min = area of smallest ball in training set
Postprocessing threshold in Algorithm 1 that rejects small detection contours; derived from the training set.
baseline dropout probability = 0.5
Selected as best single-image SweatyNet variant in Table 1 and used as the baseline in Table 2(b); model selection is not accounted for in the reported gains.

assumptions (4)

domain assumption The manual bounding-box annotations in SoccerData are correct.
All training targets and evaluation labels derive from these annotations; annotation errors would propagate directly into the reported metrics (Section 4.1).
domain assumption A binormal heatmap centered in the bounding box is an adequate supervision signal for ball detection.
The network regresses this synthetic Gaussian target rather than the bounding box directly; if the variance does not match the apparent ball size, the learned representation is distorted (Section 4.1).
domain assumption Pre-training on synthetically generated ball trajectories transfers to real RoboCup footage.
The authors use synthetic pre-training because real sequential data are scarce, and the comparison of ft versus real variants in Table 2(b) is the only indirect evidence for transfer (Section 4.1).
domain assumption The single-frame baseline and the temporal models are evaluated on the same held-out test set, with no temporal fine-tuning frames overlapping that test set.
The temporal train/test split is never stated, and the temporal models are fine-tuned on real RoboCup 2018 sequences before Table 2(b) is reported; if the test sets differ or overlap, the head-to-head comparison is invalid (Section 4.1, Section 5).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Utilizing Temporal Information in Deep Convolutional Network for Efficient Soccer Ball Detection and Tracking." pith.science (2026). https://pith.science/paper/VHORBQQI

@misc{pith2026190902406,
  author       = {Pith},
  title        = {Pith review of: Utilizing Temporal Information in Deep Convolutional Network for Efficient Soccer Ball Detection and Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VHORBQQI}},
  note         = {Machine review of arXiv:1909.02406}
}

read the original abstract

Soccer ball detection is identified as one of the critical challenges in the RoboCup competition. It requires an efficient vision system capable of handling the task of detection with high precision and recall and providing robust and low inference time. In this work, we present a novel convolutional neural network (CNN) approach to detect the soccer ball in an image sequence. In contrast to the existing methods where only the current frame or an image is used for the detection, we make use of the history of frames. Using history allows to efficiently track the ball in situations where the ball disappears or gets partially occluded in some of the frames. Our approach exploits spatio-temporal correlation and detects the ball based on the trajectory of its movements. We present our results with three convolutional methods, namely temporal convolutional networks (TCN), ConvLSTM, and ConvGRU. We first solve the detection task for an image using fully convolutional encoder-decoder architecture, and later, we use it as an input to our temporal models and jointly learn the detection task in sequences of images. We evaluate all our experiments on a novel dataset prepared as a part of this work. Furthermore, we present empirical results to support the effectiveness of using the history of the ball in challenging scenarios.

Figures

Figures reproduced from arXiv: 1909.02406 by the authors.

**Figure 1.** The proposed architecture with feed-forward and temporal parts. 1 https://github.com/AIS-Bonn/TemporalBallDetection [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** The prediction results, on the synthetically generated sequences. The network correctly predicts the future position and successfully keeps the size of slow moving ball with σ = 4 even when the history is sparse. Note that sparse history resembles an occluded ball. Input Hidden Layer Dilation =1 Output Dilation=8 Hidden Layer Dilation =4 Hidden Layer Dilation =2 x tanh σ σ tanh σ x x + * * forget gate cell state inp… view at source ↗

**Figure 3.** Visualization of a) a stack of causal convolutional layers which compose TCN architecture. b) a convolutional LSTM cell. Our approach illustrated in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: The result of the temporal part, trained on a dataset with one ball per frame. Note that the network can generalize to detect two moving objects. a) b) c) d) e) [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Qualitative results of the trained network in detecting two balls. a) SweatyNet prediction b) residual information c) ground truth d) temporal prediction e) real image images are extracted from a video recorded from the robot’s point of view and are manually annotated …

**Figure 6.** Figure 6: Top row is the part of the input history (frame {18,19,20}). The bottom row consists of heatmaps where a) visualization of the residual information from Sweaty Net to temporal, b) ground truth ball position and c) predicted output by the temporal part [PITH_FULL_IMAGE…

**Figure 7.** Figure 7: Example of correctly detected ball after finetuning with the temporal model while the confidence of just the SweatyNet is very low, resulting in false negative detection. The left image is the real image; the middle is SweatyNet output without finetuning; the right one…

**Figure 8.** Figure 8: From left to right: input image, the ground truth, prediction of the neural network, and the final output after post-processing [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 16 canonical work pages

[1]

IEEE transactions on pat- tern analysis and machine intelligence 39(12), 2481–2495 (2017)

Badrinarayanan, V., Kendall, A., Cipolla, R.: Segnet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE transactions on pat- tern analysis and machine intelligence 39(12), 2481–2495 (2017)

2017
[2]

arXiv preprint arXiv:1803.01271 (2018)

Bai, S., Kolter, J.Z., Koltun, V.: An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271 (2018)

arXiv 2018
[3]

In: ICLR (2016)

Ballas, N., Yao, L., Pal, C., Courville, A.: Delving deeper into convolutional net- works for learning video representations. In: ICLR (2016)

work page 2016
[4]

In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Carreira, J., Zisserman, A.: Quo vadis, action recognition? a new model and the kinetics dataset. In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 6299–6308 (2017)

2017
[5]

arXiv preprint arXiv:1412.3555 (2014)

Chung, J., Gulcehre, C., Cho, K., Bengio, Y.: Empirical evaluation of gated recur- rent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555 (2014)

arXiv 2014
[6]

Deep Learning for Semantic Segmentation on Minimal Hardware

van Dijk, S.G., Scheunemann, M.M.: Deep learning for semantic segmentation on minimal hardware. arXiv preprint arXiv:1807.05597 (2018)

work page Pith review arXiv 2018
[7]

In: Proceedings of 10th Workshop on Hu- manoid Soccer Robots, IEEE-RAS Int

Farazi, H., Allgeuer, P., Behnke, S.: A monocular vision system for playing soccer in low color information environments. In: Proceedings of 10th Workshop on Hu- manoid Soccer Robots, IEEE-RAS Int. Conference on Humanoid Robots. Seoul, Korea (2015)

work page 2015
[8]

Farazi, H., Karimi, M., Hosseini, M., Jegarian, M., Rahmati, D., Aghaeizadeh, F., Sadeghnejad, S., Ghidary, S.S.: Autman kid-size team description 2013. Tech. rep., Amirkabir University of Technology (2013)

work page 2013

Show all 26 references

[9]

In: 2018 IEEE-RAS 18th International Conference on Humanoid Robots (Humanoids)

Ficht, G., Farazi, H., Brandenburger, A., Rodriguez, D., Pavlichenko, D., Allgeuer, P., Hosseini, M., Behnke, S.: Nimbro-op2x: Adult-sized open-source 3d printed humanoid robot. In: 2018 IEEE-RAS 18th International Conference on Humanoid Robots (Humanoids). pp. 1–9. IEEE (2018)

2018
[10]

In: European Symposium on Artiﬁcial Neural Networks, Computational Intelligence and Machine Learning (ESANN)

Hafez Farazi, Behnke, S.: Frequency domain transformer networks for video pre- diction. In: European Symposium on Artiﬁcial Neural Networks, Computational Intelligence and Machine Learning (ESANN). Bruges, Belgium (2019)

2019
[11]

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

Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural computation 9(8), 1735–1780 (1997)

1997
[12]

arXiv preprint arXiv:1807.08405 (2018)

Houliston, T., Chalup, S.K.: Visual mesh: Real-time object detection using con- stant sample density. arXiv preprint arXiv:1807.08405 (2018)

2018 arXiv
[13]

In: Robot World Cup

Javadi, M., Azar, S.M., Azami, S., Ghidary, S.S., Sadeghnejad, S., Baltes, J.: Hu- manoid robot detection using deep learning: a speed-accuracy tradeoﬀ. In: Robot World Cup. pp. 338–349. Springer (2017)

2017
[14]

arXiv preprint arXiv:1412.6980 (2014) 14 A

Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014) 14 A. Kukleva, A. Khan et al

2014 arXiv
[15]

In: Proceedings of the ﬁrst international conference on Au- tonomous agents

Kitano, H., Asada, M., Kuniyoshi, Y., Noda, I., Osawa, E.: Robocup: The robot world cup initiative. In: Proceedings of the ﬁrst international conference on Au- tonomous agents. pp. 340–347. ACM (1997)

1997
[16]

In: proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition

Lea, C., Flynn, M.D., Vidal, R., Reiter, A., Hager, G.D.: Temporal convolutional networks for action segmentation and detection. In: proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition. pp. 156–165 (2017)

2017
[17]

arXiv preprint arXiv:1811.12493 (2018)

Leiva, F., Cruz, N., Bugue˜ no, I., Ruiz-del Solar, J.: Playing soccer without colors in the spl: A convolutional neural network approach. arXiv preprint arXiv:1811.12493 (2018)

2018 arXiv
[18]

In: 3D Vision (3DV), 2016 Fourth International Conference on

Milletari, F., Navab, N., Ahmadi, S.A.: V-net: Fully convolutional neural networks for volumetric medical image segmentation. In: 3D Vision (3DV), 2016 Fourth International Conference on. pp. 565–571. IEEE (2016)

2016
[19]

Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., Lerer, A.: Automatic diﬀerentiation in pytorch (2017)

2017
[20]

In: International Conference on Medical image computing and computer-assisted intervention

Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedi- cal image segmentation. In: International Conference on Medical image computing and computer-assisted intervention. pp. 234–241. Springer (2015)

2015
[21]

In: Proceedings of the 12th Workshop on Humanoid Soccer Robots, IEEE-RAS Inter- national Conference on Humanoid Robots, Birmingham (2017)

Schnekenburger, F., Scharﬀenberg, M., W¨ ulker, M., Hochberg, U., Dorer, K.: De- tection and localization of features on a soccer ﬁeld with feedforward fully con- volutional neural networks (fcnn) for the adultsize humanoid robot sweaty. In: Proceedings of the 12th Workshop on...

2017
[22]

In: Proc

Schulz, H., Strasdat, H., Behnke, S.: A ball is not just orange: Using color and lu- minance to classify regions of interest. In: Proc. of Second Workshop on Humanoid Soccer Robots, Pittsburgh (2007)

2007
[23]

In: Robot World Cup

Speck, D., Barros, P., Weber, C., Wermter, S.: Ball localization for robocup soccer using convolutional neural networks. In: Robot World Cup. pp. 19–30. Springer (2016)

2016
[24]

The Journal of Machine Learning Research 15(1), 1929–1958 (2014)

Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., Salakhutdinov, R.: Dropout: a simple way to prevent neural networks from overﬁtting. The Journal of Machine Learning Research 15(1), 1929–1958 (2014)

2014
[25]

International journal of computer vision 103(1), 60–79 (2013)

Wang, H., Kl¨ aser, A., Schmid, C., Liu, C.L.: Dense trajectories and motion bound- ary descriptors for action recognition. International journal of computer vision 103(1), 60–79 (2013)

2013
[26]

In: Advances in neural information processing systems

Xingjian, S., Chen, Z., Wang, H., Yeung, D.Y., Wong, W.K., Woo, W.c.: Convo- lutional lstm network: A machine learning approach for precipitation nowcasting. In: Advances in neural information processing systems. pp. 802–810 (2015)

2015

Pith tools

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

[1] [1]

IEEE transactions on pat- tern analysis and machine intelligence 39(12), 2481–2495 (2017)

Badrinarayanan, V., Kendall, A., Cipolla, R.: Segnet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE transactions on pat- tern analysis and machine intelligence 39(12), 2481–2495 (2017)

2017

[2] [2]

arXiv preprint arXiv:1803.01271 (2018)

Bai, S., Kolter, J.Z., Koltun, V.: An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271 (2018)

arXiv 2018

[3] [3]

In: ICLR (2016)

Ballas, N., Yao, L., Pal, C., Courville, A.: Delving deeper into convolutional net- works for learning video representations. In: ICLR (2016)

work page 2016

[4] [4]

In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Carreira, J., Zisserman, A.: Quo vadis, action recognition? a new model and the kinetics dataset. In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 6299–6308 (2017)

2017

[5] [5]

arXiv preprint arXiv:1412.3555 (2014)

Chung, J., Gulcehre, C., Cho, K., Bengio, Y.: Empirical evaluation of gated recur- rent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555 (2014)

arXiv 2014

[6] [6]

Deep Learning for Semantic Segmentation on Minimal Hardware

van Dijk, S.G., Scheunemann, M.M.: Deep learning for semantic segmentation on minimal hardware. arXiv preprint arXiv:1807.05597 (2018)

work page Pith review arXiv 2018

[7] [7]

In: Proceedings of 10th Workshop on Hu- manoid Soccer Robots, IEEE-RAS Int

Farazi, H., Allgeuer, P., Behnke, S.: A monocular vision system for playing soccer in low color information environments. In: Proceedings of 10th Workshop on Hu- manoid Soccer Robots, IEEE-RAS Int. Conference on Humanoid Robots. Seoul, Korea (2015)

work page 2015

[8] [8]

Farazi, H., Karimi, M., Hosseini, M., Jegarian, M., Rahmati, D., Aghaeizadeh, F., Sadeghnejad, S., Ghidary, S.S.: Autman kid-size team description 2013. Tech. rep., Amirkabir University of Technology (2013)

work page 2013