Trajectory Prediction by Coupling Scene-LSTM with Human Movement LSTM

Gita Alaghband; Manh Huynh

REVIEW 4 major objections 4 minor 23 references

Trajectory Prediction by Coupling Scene-LSTM with Human Movement LSTM

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Scene-LSTM couples a two-level scene grid of common paths with per-pedestrian LSTM motion to beat social-interaction models on five benchmark datasets.

desk verdict The scene-memory idea is real, but the quantitative story collapses when you notice the same LSTM model is reported as 0.66 and 1.74 ADE across two tables. read the letter →

arxiv 1908.08908 v1 pith:2RMQSWAR submitted 2019-08-23 cs.CV

Manh Huynh , Gita Alaghband This is my paper · ORCID

classification cs.CV

keywords humantrajectorypredictionLSTMsceneinformationcommonmovementtwo-levelgridhardfiltersoftcrowdanalysis

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper sets out to show that the next positions of a pedestrian in a static crowd can be predicted more accurately by learning the common paths people already take through each local part of the scene than by modeling social interactions alone. The authors build a two-level grid over the video frame, train a Scene-LSTM for every grid cell to absorb all observed movements in that cell, and train a separate Pedestrian-LSTM for each person's own relative motion. A hard filter restricts the scene memory to non-linear cells and to subgrids crossed by at least three people, and a soft filter multiplies the scene memory by a gating vector derived from the pedestrian's location and recent motion. On ETH, UCY, Town Center, PETS09, and Grand Central they report average displacement error falling from 0.79 to 0.56 meters and final displacement error from 1.65 to 1.15 meters compared with the best method they compare against. If the claim holds, recurrent scene-specific structure is a cheaper and stronger cue than the interaction graphs most current predictors rely on.

What carries the argument

The load-bearing mechanism is the two-level grid paired with the Scene Data Filter. Each grid cell carries a Scene-LSTM whose memory represents how people move through that cell; a cell is labeled linear if all training trajectories inside it are straight, and in linear cells the scene memory is ignored because the Pedestrian-LSTM alone suffices. For non-linear cells, an $m\times m$ subgrid parses each trajectory into subgrid paths, and paths traveled by more than $p=3$ pedestrians define 'common subgrids' that the hard filter treats as trustworthy. A soft filter then gates the selected cell memory with a sigmoid vector derived from the pedestrian's current subgrid and LSTM state, so the scene influence is tuned to the individual's recent behavior. The filtered scene vector is added to the pedestrian hidden state before a bivariate Gaussian is sampled for the next displacement; this coupling is what lets common local motion steer predictions without contaminating them with rare trajectories.

What would settle it

Swap the training and testing halves of each video in Stage 2 (train on the second 50% of frames, test on the first 50%). If the ADE and FDE gains over the Pedestrian-LSTM-only model shrink or disappear, then the hard filter is exploiting the particular half of the video rather than a stable property of the scene.

Watch

Extended reading notes

Core claim

The central claim is that a static scene's commonly traveled paths, encoded in a grid of LSTM cells, are sufficient to beat models built on social interaction. Each grid cell's Scene-LSTM is updated with the concatenation of a one-hot vector marking the pedestrian's subgrid position and that pedestrian's LSTM hidden state, so the cell memory accumulates typical local motion. The hard filter classifies each cell as linear or non-linear at preprocessing time; only non-linear cells are kept, their trajectories are parsed into subgrid paths, and only subgrids shared by more than $p=3$ pedestrians count as common. At prediction time, a pedestrian whose current subgrid is common receives the cell's hidden state, and the soft filter element-wise multiplies that state with a sigmoid gate computed from the one-hot location and the pedestrian hidden state, then adds the result into the pedestrian hidden state before sampling the next displacement from a bivariate Gaussian. The authors report that this combined model reduces average ADE from 0.79 to 0.56 meters and FDE from 1.65 to 1.15 meters on the five ETH/UCY sequences, and outperforms the same comparison methods on Town Center, PETS09, and Grand Central.

Load-bearing premise

The load-bearing premise is that the common paths learned from the first half of a video's frames remain valid for the second half, so a pedestrian standing in a common subgrid can safely be steered by that learned path.

Editorial extensions

If this is right

If the reported gains are real, trajectory prediction in static crowded scenes can rely primarily on scene-specific common paths, reducing the need for expensive social-interaction modeling.
The two-level grid and hard/soft filtering can be recomputed from raw trajectories alone, so the method should transfer to a new static scene after a short re-training pass on half of the new video's frames.
The ablation results imply that subgrid granularity is the controlling factor: too coarse a grid merges distinct paths, too fine a grid fragments them, and an 8x8 cell/subgrid split best captures common motion.
For robotics and tracking applications that require one best-guess trajectory, the model's single sampled prediction with lower FDE means better long-horizon endpoint estimates than the compared single-trajectory baselines.
On the three additional datasets, the improvements hold at longer horizons (6.4 seconds) as well as shorter ones, so the common-path signal is not only a short-range effect.

Reading between the lines

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

A natural stress test is to make the hard filter time-aware: recompute common subgrids over a sliding window, which would let the model follow shifts in crowd flow after events or train arrivals rather than freezing the first half of the video.
The same grid-plus-filter recipe could be applied to other agents, such as vehicles, cyclists, or warehouse robots, wherever recurring local paths dominate motion; the $p=3$ threshold would become a per-scene hyperparameter.
The dependency on stage-2 training suggests an implicit test: if the scene memory is trained on a different time of day or on a disjoint set of pedestrians and the gain shrinks, the 'common' paths are partly person-specific rather than purely scene-specific.
Combining the soft-filtered scene memory with a social pooling layer looks like the natural next architecture; the ablation's NDE numbers suggest the grid-level scene memory and the subgrid-level filter capture complementary parts of the error.

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

You should know this paper for the architecture, not for the numbers. The Scene-LSTM idea — a two-level grid that learns common routes from observed trajectories, plus a hard filter for linear/non-linear regions and a soft filter to gate the scene memory per pedestrian — is genuinely interesting and not something I've seen in earlier scene-aware trajectory work. The ablation study suggests the components do something: PMrel alone gets 0.66 ADE, and the full model gets 0.56 on ETH/UCY. That 0.1 m gain is plausible and worth investigating.

The problem is that the quantitative comparison to prior methods cannot be trusted. The LSTM baseline in Table 1 is reported at 1.74 ADE, but the PMrel ablation in Table 2 is 0.66 ADE. These are the same model: a per-pedestrian LSTM with relative displacements and Gaussian output. Under the stated protocol ("we apply the same training and testing procedures for all methods"), they should be identical. They are not. This means the Table 1 baseline numbers were probably taken from earlier papers rather than rerun under the Stage 2 fine-tuning protocol. If the baselines weren't rerun, the comparison is uncontrolled, and the headline reduction from 0.79 to 0.56 ADE over SGAN-20V-1 does not support the claim that the scene memory is responsible. The more modest comparison in the ablation — 0.66 vs 0.56 — is the only controlled one, and it still supports the method, but with a much smaller effect.

Two other soft spots. The grid and subgrid sizes are tuned on ETH/UCY and then tested on ETH/UCY, which can inflate the result. And the Stage 2 fine-tuning on 50% of the target video is a strong protocol, but unless every baseline gets the same fine-tuning the comparison is not fair. No error bars or code either.

Who should read this: anyone designing scene-conditioned prediction models will find the filtering mechanism worth thinking about. But the paper should be cited cautiously, if at all, until the baselines are rerun.

Recommendation: send it to review, but insist on a controlled re-evaluation of all baselines under the same protocol, error bars, and code. If the authors do that, the core result might hold up, but at a smaller scale than claimed.

Referee Report

4 major / 4 minor

Summary. The paper proposes Scene-LSTM, a trajectory-prediction model that couples a per-pedestrian LSTM (Pedestrian-LSTM) with a scene-level LSTM (Scene-LSTM) operating on a two-level grid. A hard filter decides whether to use scene information based on whether the pedestrian is in a linear grid cell or a common subgrid path, and a soft filter selects the relevant scene memory via an attention-like gating mechanism. The model is evaluated on ETH, UCY, Town Center, PETS09, and Grand Central, reporting lower average ADE and FDE than baselines including Linear, LSTM, Social-LSTM, and SGAN variants. An ablation study isolates the contributions of relative coordinates, scene data, hard filter at grid/subgrid levels, and the soft filter.

Significance. If the quantitative claims are substantiated, the two-level grid scene memory with hard/soft filtering is a simple and interpretable alternative to visual-feature-based scene models, and the additional evaluation on Town Center, PETS09, and Grand Central is a useful generalization check. The paper's strengths include a clear problem formulation, a systematic ablation of PMabs vs PMrel, and explicit consideration of when scene data should or should not be used. However, the reported numerical comparisons are undermined by an internal inconsistency between the LSTM baseline in Table 1 and the identical PMrel model in Table 2, and by the absence of controlled hyperparameter selection, error bars, and significance tests. The central claim of a large improvement therefore is not yet supported as stated.

major comments (4)

[§5.2, Table 2 vs §5.1, Table 1] PMrel in Table 2 is a per-pedestrian LSTM with relative displacement input and Gaussian output, which is the same model as the LSTM baseline in Table 1. Under the paper's stated protocol ('we apply the same training and testing procedures for all methods'), these numbers should agree or nearly agree, but they differ by 1.08 m ADE (0.66 vs 1.74) and 1.81 m FDE (1.36 vs 3.17). This discrepancy indicates that the Table 1 baselines were either not run under the same two-stage fine-tuning protocol or were taken from prior publications. Since PMrel alone comes within 0.10 ADE and 0.21 FDE of the full Scene-LSTM, the headline improvement from 0.79 to 0.56 ADE cannot be cleanly attributed to the scene-memory mechanism. The authors must rerun all baselines under the identical Stage 1 plus Stage 2 protocol and report consistent numbers.
[§5.1, §5.4, Figure 5] The grid and subgrid sizes (selected as 8x8 in Figure 5b) and the stage-2 training fraction of 50% are tuned on the same five ETH/UCY videos used for the main results, with no held-out validation split. This selection on the test data can inflate reported gains. Additionally, no error bars, multiple seeds, or significance tests are reported, so the phrase 'significantly outperforms' is not statistically supported. Please fix hyperparameters using a separate validation set or nested cross-validation, and report variance and significance measures.
[§4, Implementation Details, Hard filter] The hard filter's core decisions depend on unspecified algorithmic criteria: 'a grid cell is characterized as linear if all human trajectories in this grid cell are linear' and 'a common subgrid path is a path between two subgrids commonly travelled by a number of pedestrians greater than p = 3'. No precise definition of linearity (e.g., curvature threshold, number of points, fitting residual) or of how a trajectory is parsed into subgrid paths is provided. These details are load-bearing because they determine which scene data are used at prediction time, and without them the method cannot be reproduced or compared fairly.
[§5.1, Table 1, NDE results] The average NDE advantage over SGAN-20V-1 (1.00 vs 1.36) is not consistent across sequences: on ETH-Univ Scene-LSTM is worse (1.88 vs 1.55) and on UCY-Zara02 it is also worse (0.93 vs 0.88). The paper acknowledges this qualitatively, but reporting only sequence-level numbers without error bars or significance tests leaves it unclear whether the average difference is meaningful. Per-sequence confidence intervals or a paired test over multiple seeds would clarify this.

minor comments (4)

[Abstract and author affiliation] The abstract contains 'a pedestrians future locations' missing an apostrophe, and the author affiliation contains 'Univesity' and later 'ﬁler' in the contributions list.
[§5.4] The reported prediction horizons 'Tpred = 4.8 and 6.4 seconds' are not tied to the observed-frame length or the frame rates of Town Center, PETS09S2, and Grand Central; please specify the number of frames and frame rates for each dataset.
[Figure 5] The red and blue lines mentioned in the text are not labeled directly in the figure; please add a legend and axis labels so the reader can interpret the grid/subgrid-size experiments.
[§5, Metrics] The metrics are described as 'mean square error (Euclidean distance)', which is inaccurate because the reported values are average Euclidean distances, not squared errors; please rephrase to 'average Euclidean distance' or 'mean squared Euclidean distance' consistently.

Circularity Check

0 steps flagged · score 0.0 of 10

No meaningful circularity: Scene-LSTM predictions are evaluated on held-out frames, and no equation-level reduction of the reported predictions to fitted inputs or self-citations was found.

full rationale

The paper's derivation is self-contained in the sense required by the circularity check. The Scene-LSTM common-path model is trained on observed frames, and predictions are evaluated on remaining held-out frames. The hard filter's linear/non-linear labels and common subgrids are computed from the training portion of each video only, while the Scene-LSTM and Pedestrian-LSTM parameters are optimized by negative log-likelihood on training trajectories, not on the test trajectories. The two-stage protocol (leave-one-out pretraining followed by 10 epochs on 50% of the target video) is stated explicitly and, according to the paper, applied to all compared methods. The choice of grid and subgrid sizes via the benchmark error curves is a hyperparameter-selection procedure on the same datasets; this raises an overfitting-to-benchmark concern, but it is not an equation-level identity between an input and a reported prediction. The apparent inconsistency between PMrel (0.66 ADE) in Table 2 and the LSTM baseline (1.74 ADE) in Table 1 is a serious reporting or comparison problem, but it concerns whether baseline numbers were obtained under the same protocol, not a reduction of the central claim to its own inputs. No load-bearing self-citation or imported uniqueness theorem was found; the cited prior works are standard baselines and architecture sources, and the claimed improvements are not derived from those citations by construction.

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

The central claim rests on several free hyperparameters and heuristic domain assumptions about pedestrian behavior and scene stability. None of these are derived from first principles; they are chosen by hand or tuned on the evaluation datasets, which adds to the circularity burden.

free parameters (5)

grid_size = 8x8
Selected by sweeping grid sizes and picking the one with lowest ADE/FDE on the ETH/UCY datasets (Figure 5b).
subgrid_size = 8x8
Selected in the same sweep as grid size; 8x8 was best on the evaluation datasets (Figure 5b).
common_path_threshold_p = 3
Hand-chosen threshold to define a 'common' subgrid path: a path travelled by more than 3 pedestrians (Section 3(c)). No sensitivity analysis is provided.
hidden_state_size = 128
All memory cells and hidden state vectors are set to 128 (Section 5, Implementation details).
stage2_training_fraction = 0.5
The best model is fine-tuned on 50% of the target video frames (Stage 2); an experiment shows performance improves with more stage-2 data (Figure 5a).

assumptions (4)

domain assumption Grid cells with only linear trajectories are not affected by scene structure, so scene data should be ignored for them.
Introduced in Section 3(c) and Figure 3; separates linear and non-linear grid cells without a formal criterion.
domain assumption Pedestrians located in a common subgrid tend to follow the common path; scene memory helps them.
Core heuristic of the hard filter at subgrid level (Section 3(c), Figure 3f).
domain assumption Common movement patterns learned from the first 50% of a video remain valid in the remaining 50% used for testing.
Implied by the Stage 2 training protocol (Section 5, Training).
standard math The bivariate Gaussian output distribution and negative log-likelihood loss are appropriate for trajectory prediction.
Adopted from Social-LSTM [1] in Equations 8-11.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Trajectory Prediction by Coupling Scene-LSTM with Human Movement LSTM." pith.science (2026). https://pith.science/paper/2RMQSWAR

@misc{pith2026190808908,
  author       = {Pith},
  title        = {Pith review of: Trajectory Prediction by Coupling Scene-LSTM with Human Movement LSTM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2RMQSWAR}},
  note         = {Machine review of arXiv:1908.08908}
}

read the original abstract

We develop a novel human trajectory prediction system that incorporates the scene information (Scene-LSTM) as well as individual pedestrian movement (Pedestrian-LSTM) trained simultaneously within static crowded scenes. We superimpose a two-level grid structure (grid cells and subgrids) on the scene to encode spatial granularity plus common human movements. The Scene-LSTM captures the commonly traveled paths that can be used to significantly influence the accuracy of human trajectory prediction in local areas (i.e. grid cells). We further design scene data filters, consisting of a hard filter and a soft filter, to select the relevant scene information in a local region when necessary and combine it with Pedestrian-LSTM for forecasting a pedestrian's future locations. The experimental results on several publicly available datasets demonstrate that our method outperforms related works and can produce more accurate predicted trajectories in different scene contexts.

Figures

Figures reproduced from arXiv: 1908.08908 by the authors.

**Figure 1.** Scene-LSTM learns common human movements on a two-level grid structure. The common human movement is filtered and used in combination with individual movement (PedestrianLSTM) to predict a pedestrians future locations. To partially handle this multi-modal problem, the contexts of a scene can be used. Several proposed methods [21,2,17,3] have gained improvements by extracting the scenes visual features from video… view at source ↗

**Figure 2.** The system consists of three main modules: Pedestrian Movement (PM), Scene Data (SD) and Scene Data Filter (SDF). PM models the individual movement of pedestrians. SD encodes common human movements in each grid cell. SDF selects relevant scene data to update the Pedestrian-LSTM, which is used to predict the future locations. ⊗ denotes elementwise multiplication. ⊕ denotes vector addition. hi and hs are the hidden s… view at source ↗

**Figure 3.** Illustrations of the hard filter, which determines whether the scene data should be applied in predicting the future locations of a pedestrian. (a) the frame image is first divided into n × n grid cells (n = 4 in this example) to capture all human movements in each grid cell; (b) & (c) only non-linear grid cells are selected for further processing at the subgrid level; the scene data is not applied for pedestrians i… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Illustrations of the soft filter. The relevant information of scene data (i.e. Scene-LSTM) is selected using each pedestrians walking behavior. The filtered grid-cell memory of each pedestrian is then used in combination with pedestrian movements (Pedestrian-LSTM) to …

**Figure 5.** Figure 5: the impact of (a) training data amount at stage-2 and (b) finding grid/subgrid sizes to optimize prediction accuracy. The results (ADE and FDE) are calculated on average of 5 videos [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

**Figure 6.** Figure 6: Qualitative comparison between our method with SGAN-20VP-1 in different scenarios: (a) pedestrians are standing still and waiting for trains, (b) a pedestrian is entering the door, (c) a pedestrian is finding a path between obstacle and trees, (d) a pedestrian makes a …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 20 canonical work pages

[1]

In: Proceedings of the IEEE conference on computer vision and pattern recognition

Alahi, A., Goel, K., Ramanathan, V ., Robicquet, A., Fei-Fei, L., Savarese, S.: Social lstm: Human trajectory prediction in crowded spaces. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 961–971 (2016)

work page 2016
[2]

In: European Conference on Computer Vision

Ballan, L., Castaldo, F., Alahi, A., Palmieri, F., Savarese, S.: Knowledge transfer for scene- speciﬁc motion prediction. In: European Conference on Computer Vision. pp. 697–713. Springer (2016)

work page 2016
[3]

In: International Conference on Pattern Recognition

Bartoli, F., Lisanti, G., Ballan, L., Del Bimbo, A.: Context-aware trajectory prediction. In: International Conference on Pattern Recognition. pp. 1941–1946. IEEE (2018)

work page 2018
[4]

In: CVPR

Benfold, B., Reid, I.: Stable multi-target tracking in real-time surveillance video. In: CVPR. pp. 3457–3464. IEEE (2011)

work page 2011
[5]

In: Twelfth IEEE International Workshop on Performance Evaluation of Tracking and Surveillance

Ferryman, J., Shahrokni, A.: Pets2009: Dataset and challenge. In: Twelfth IEEE International Workshop on Performance Evaluation of Tracking and Surveillance. pp. 1–6. IEEE (2009) 14 Manh Huynh and Gita Alaghband

work page 2009
[6]

In: Advances in neural information processing systems

Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y .: Generative adversarial nets. In: Advances in neural information processing systems. pp. 2672–2680 (2014)

2014
[7]

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

Gupta, A., Johnson, J., Fei-Fei, L., Savarese, S., Alahi, A.: Social gan: Socially acceptable trajectories with generative adversarial networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 2255–2264 (2018)

work page 2018
[8]

arXiv preprint arXiv:1412.6980 (2014)

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

arXiv 2014

Show all 23 references

[9]

In: 29th IEEE Conference on Decision and Control

Leonard, J.J., Durrant-Whyte, H.F.: Application of multi-target tracking to sonar-based mo- bile robot navigation. In: 29th IEEE Conference on Decision and Control. pp. 3118–3123. IEEE (1990)

1990
[10]

In: Computer graphics forum

Lerner, A., Chrysanthou, Y ., Lischinski, D.: Crowds by example. In: Computer graphics forum. vol. 26, pp. 655–664. Wiley Online Library (2007)

2007
[11]

In: 2011 IEEE Intelligent Vehicles Symposium (IV)

Levinson, J., Askeland, J., Becker, J., Dolson, J., Held, D., Kammel, S., Kolter, J.Z., Langer, D., Pink, O., Pratt, V ., et al.: Towards fully autonomous driving: Systems and algorithms. In: 2011 IEEE Intelligent Vehicles Symposium (IV). pp. 163–168. IEEE (2011)

2011
[12]

In: 2018 15th Conference on Computer and Robot Vision (CRV)

Manh, H., Alaghband, G.: Spatiotemporal ksvd dictionary learning for online multi-target tracking. In: 2018 15th Conference on Computer and Robot Vision (CRV). pp. 150–157. IEEE (2018)

2018
[13]

PyTorch: Tensors and dynamic neural net- works in Python with strong GPU acceleration 6 (2017)

Paszke, A., Gross, S., Chintala, S., Chanan, G.: Pytorch: Tensors and dynamic neural net- works in python with strong gpu acceleration. PyTorch: Tensors and dynamic neural net- works in Python with strong GPU acceleration 6 (2017)

2017
[14]

In: Computer Vision, 2009 IEEE 12th International Conference on

Pellegrini, S., Ess, A., Schindler, K., Van Gool, L.: You’ll never walk alone: Modeling so- cial behavior for multi-target tracking. In: Computer Vision, 2009 IEEE 12th International Conference on. pp. 261–268. IEEE (2009)

2009
[15]

In: European conference on computer vision

Robicquet, A., Sadeghian, A., Alahi, A., Savarese, S.: Learning social etiquette: Human tra- jectory understanding in crowded scenes. In: European conference on computer vision. pp. 549–565. Springer (2016)

2016
[16]

In: 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems

Trautman, P., Krause, A.: Unfreezing the robot: Navigation in dense, interacting crowds. In: 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 797–803. IEEE (2010)

2010
[17]

arXiv preprint arXiv:1705.09436 (2017)

Varshneya, D., Srinivasaraghavan, G.: Human trajectory prediction using spatially aware deep attention models. arXiv preprint arXiv:1705.09436 (2017)

2017 arXiv
[18]

In: IEEE International Conference on Robotics and Automation (ICRA)

Vemula, A., Muelling, K., Oh, J.: Social attention: Modeling attention in human crowds. In: IEEE International Conference on Robotics and Automation (ICRA). pp. 1–7. IEEE (2018)

2018
[19]

IEEE Transactions on Intelligent Transportation Systems 19(2), 582–597 (2017)

Vivacqua, R.P.D., Bertozzi, M., Cerri, P., Martins, F.N., Vassallo, R.F.: Self-localization based on visual lane marking maps: An accurate low-cost approach for autonomous driv- ing. IEEE Transactions on Intelligent Transportation Systems 19(2), 582–597 (2017)

2017
[20]

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

Xu, Y ., Piao, Z., Gao, S.: Encoding crowd interaction with deep neural network for pedes- trian trajectory prediction. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 5275–5284 (2018)

2018
[21]

In: 2018 IEEE Winter Conference on Applications of Computer Vision (W ACV)

Xue, H., Huynh, D.Q., Reynolds, M.: Ss-lstm: a hierarchical lstm model for pedestrian tra- jectory prediction. In: 2018 IEEE Winter Conference on Applications of Computer Vision (W ACV). pp. 1186–1194. IEEE (2018)

2018
[22]

Yamaguchi, K., Berg, A.C., Ortiz, L.E., Berg, T.L.: Who are you with and where are you going? In: CVPR 2011. pp. 1345–1352. IEEE (2011)

2011
[23]

In: 2012 IEEE Conference on Computer Vision and Pattern Recognition

Zhou, B., Wang, X., Tang, X.: Understanding collective crowd behaviors: Learning a mixture model of dynamic pedestrian-agents. In: 2012 IEEE Conference on Computer Vision and Pattern Recognition. pp. 2871–2878. IEEE (2012)

2012

Pith tools

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

[1] [1]

In: Proceedings of the IEEE conference on computer vision and pattern recognition

Alahi, A., Goel, K., Ramanathan, V ., Robicquet, A., Fei-Fei, L., Savarese, S.: Social lstm: Human trajectory prediction in crowded spaces. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 961–971 (2016)

work page 2016

[2] [2]

In: European Conference on Computer Vision

Ballan, L., Castaldo, F., Alahi, A., Palmieri, F., Savarese, S.: Knowledge transfer for scene- speciﬁc motion prediction. In: European Conference on Computer Vision. pp. 697–713. Springer (2016)

work page 2016

[3] [3]

In: International Conference on Pattern Recognition

Bartoli, F., Lisanti, G., Ballan, L., Del Bimbo, A.: Context-aware trajectory prediction. In: International Conference on Pattern Recognition. pp. 1941–1946. IEEE (2018)

work page 2018

[4] [4]

In: CVPR

Benfold, B., Reid, I.: Stable multi-target tracking in real-time surveillance video. In: CVPR. pp. 3457–3464. IEEE (2011)

work page 2011

[5] [5]

In: Twelfth IEEE International Workshop on Performance Evaluation of Tracking and Surveillance

Ferryman, J., Shahrokni, A.: Pets2009: Dataset and challenge. In: Twelfth IEEE International Workshop on Performance Evaluation of Tracking and Surveillance. pp. 1–6. IEEE (2009) 14 Manh Huynh and Gita Alaghband

work page 2009

[6] [6]

In: Advances in neural information processing systems

Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y .: Generative adversarial nets. In: Advances in neural information processing systems. pp. 2672–2680 (2014)

2014

[7] [7]

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

Gupta, A., Johnson, J., Fei-Fei, L., Savarese, S., Alahi, A.: Social gan: Socially acceptable trajectories with generative adversarial networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 2255–2264 (2018)

work page 2018

[8] [8]

arXiv preprint arXiv:1412.6980 (2014)

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

arXiv 2014