mdongbenben / OpenTraj

Trajectory Prediction Benchmark and State-of-the-art

OpenTraj

Trajectory Prediction Benchmark and State-of-the-art

Table of Public Available Trajectory Datasets

Sample	Name	#Trajs	Coord	FPS	Density	******Description******	REF
	ETH	Peds=750	world	2.5	?	2 top view scenes	[website] [paper]
	UCY	Peds=786	world	2.5	?	3 scenes (Zara/Arxiepiskopi/University). Zara and University close to top view. Arxiepiskopi more inclined.	[website] paper
	SDD	Bikes=4210 Peds=5232 Skates=292 Carts=174 Cars=316 Buss=76 Total=10,300	image	30	?	8 top view scenes	[website] [paper]
	GC	Peds=12,684	image	25	?	1 scene	[dropbox] [paper]
	Waymo			?	?		website github
	KITTI		image(3d) +Calib	10	?		website
	HERMES				?		website
	highD	> 110,500 vehicles			?		website
	inD	Vehicles= x Peds=x Bikes=x			?		website
	TRAF		image	10	?		website gDrive
	L-CAS				?		website
	VIRAT				?		website
	VRU	peds=1068 Bikes=464	World (Meter)	?	?	consists of pedestrian and cyclist trajectories, recorded at an urban intersection using cameras and LiDARs	website
	Edinburgh				?		website
	Town Center				?	1 scene	website
	ZTD	Vehicles= x	World (Degree)	10	?	ZEN Traffic Dataset: containing vehicle trajectories	website

Other Datasets

• NGSim
• Daimler
• [ATC](No Link)
• [Cyclist](No Link)

Benchmarks

• Trajnet: Trajectory Forecasting Challenge
• MOT-Challenge: Multiple Object Tracking Benchmark

Metrics

1. ADE (T_obs, T_pred): Average Displacement Error (ADE), also called Mean Euclidean Distance (MED), measures the averages Euclidean distances between points of the predicted trajectory and the ground truth that have the same temporal distance from their respective start points. The function arguemnts are:

• T_obs : observation period
• T_pred : prediction period

2. FDE (T_obs, T_pred): Final Displacement Error (FDE) measures the distance between final predicted position and the ground truth position at the corresponding time point. The function arguemnts are:

• T_obs : observation period
• T_pred : prediction period

State-of-the-arts Trajectory Prediction Algorithms

1. ETH Dataset

Method	Univ (ADE/FDE)*	Hotel (ADE/FDE)*
Social-Force 1	0.67 / 1.52	0.52 / 1.03
Social-LSTM 2	1.09 / 2.35	0.79 / 1.76
Social-GAN REF	0.77 / 1.38	0.70 / 1.43
Social-Ways REF	0.39 / 0.64	0.39 / 0.66
Social-Attention REF	?	?
SoPhie REF	?	?
CIDNN REF	?	?
Social-Etiquette REF	?	?
ConstVel REF	?	?
Scene-LSTM REF	?	?
Peeking Into the Future REF	?	?
SS-LSTM REF	?	?
MX-LSTM REF	?	?
Social-BiGAT REF	?	?
SR-LSTM REF	?	?

* The values are in meter, calculated with ADE(T_obs=3.2_s, T_pred=4.8_s) and FDE(T_obs=3.2_s, T_pred=4.8_s).

2. UCY Dataset

Method	ZARA01 (ADE/FDE)	ZARA02 (ADE/FDE)	Students (ADE/FDE)
Social-Force 1	?	?	?
Social-Etiquette REF	?	?	?
Social-LSTM 2	?	?	?
Social-GAN REF	?	?	?
CIDNN REF	?	?	?
Social-Attention REF	?	?	?
Scene-LSTM REF	?	?	?
ConstVel REF	?	?	?
SoPhie REF	?	?	?
Social-Ways REF	?	?	?
Peeking Into the Future REF	?	?	?
SS-LSTM REF	?	?	?
Social-BiGAT REF	?	?	?
SR-LSTM REF	?	?	?

3. Stanford Drone Dataset (SDD)

• Social-Etiquette
• DESIRE
• SoPhie
• MATF (Multi-Agent Tensor Fusion)
• Best of Many

4. Grand Central Station (GC) Dataset

• CIDNN

5. KITI

• R2P2

References

* ordered by time

1. Who are you with and Where are you going? (Social Force), Yamaguchi et al. CVPR 2011. paper
2. Social LSTM: Human trajectory prediction in crowded spaces, Alahi et al. CVPR 2016. paepr
3. Learning social etiquette: Human trajectory understanding in crowded scenes, Robicquet et al. ECCV 2016. paper

• Desire: Distant future prediction in dynamic scenes with interacting agents, Lee et al. CVPR 2017. paper

3. Social GAN: Socially Acceptable Trajectories with Generative Adversarial Networks, Gupta et al. CVPR 2018. paper
4. Social Ways: Learning Multi-Modal Distributions of Pedestrian Trajectories with GANs, Amirian et al. CVPR 2019. paper, code

• Sophie: An attentive gan for predicting paths compliant to social and physical constraints, Sadeghian et al. CVPR 2019. paper

• MATF (Multi-Agent Tensor Fusion)

• Best of Many

Surveys

1. A survey on motion prediction and risk assessment for intelligent vehicles, ROBOMECH 2014. paper
2. Trajectory data mining: an overview, TIST 2015. paper
3. Survey on Vision-Based Path Prediction, DAPI 2018. arxiv
4. A literature review on the prediction of pedestrian behavior in urban scenarios, ITSC 2018. paper
5. Autonomous vehicles that interact with pedestrians: A survey of theory and practice, ITS 2019. arxiv
6. Human Motion Trajectory Prediction: A Survey, IJRR 2019 arxiv
7. A Survey on Path Prediction Techniques for Vulnerable Road Users: From Traditional to Deep-Learning Approaches, ITSC 2019. paper