Simple Online and Realtime Tracking with a Deep Association Metric

Wojke, Nicolai; Bewley, Alex; Paulus, Dietrich

doi:10.48550/arxiv.1703.07402

Cited by 41 publications

(57 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An autonomous vehicle, for instance, must predict the states of objects immediately when new detections become available. Most recent approaches to multi-object tracking are therefore online methods that do not depend on future frames [3,4,7,17,25,33,45,48,52,55]. Most online methods estimate similarity scores between the detections and the existing tracks based on various cues, such as predicted bounding boxes [4] and appearance similarity [18].…”

Section: Related Workmentioning

confidence: 99%

“…While some methods [4,18] only take the last detection corresponding to a track into account, some techniques aggregate temporal information into a track history. For instance, DEEP Sort [48] computes the maximum similarity between the detection and any detections that were previously associated with the track. Other methods rely on recurrent neural networks to accumulate temporal information [7,17,26,33,55].…”

Section: Related Workmentioning

confidence: 99%

“…Data association in most existing methods is based on similarity scores computed between the detections and the existing tracks. The representation of existing tracks can rely on the last detections [48] or it can be aggregated from historical detections associated with the tracks [26,33]. Figure 2 groups approaches to multi-object tracking by how they leverage context information.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SoDA: Multi-Object Tracking with Soft Data Association

Hung,

Kretzschmar,

Lin

et al. 2020

Preprint

View full text Add to dashboard Cite

Robust multi-object tracking (MOT) is a prerequisite for a safe deployment of self-driving cars. Tracking objects, however, remains a highly challenging problem, especially in cluttered autonomous driving scenes in which objects tend to interact with each other in complex ways and frequently get occluded. We propose a novel approach to MOT that uses attention to compute track embeddings that encode the spatiotemporal dependencies between observed objects. This attention measurement encoding allows our model to relax hard data associations, which may lead to unrecoverable errors. Instead, our model aggregates information from all object detections via soft data associations. The resulting latent space representation allows our model to learn to reason about occlusions in a holistic data-driven way and maintain track estimates for objects even when they are occluded. Our experimental results on the Waymo Open Dataset suggest that our approach leverages modern largescale datasets and performs favorably compared to the state of the art in visual multi-object tracking.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SoDA: Multi-Object Tracking with Soft Data Association

Hung,

Kretzschmar,

Lin

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…FIT Dataset. We collected 18 scenes from different locations in Europe and relied on MaskRCNN [22] and deepsort [58] to detect and track objects, and DSO [18] to esti-mate the egomotion. This dataset allows testing the robustness to noisy inputs (without human annotation).…”

Section: Datasetsmentioning

confidence: 99%

Multimodal Future Localization and Emergence Prediction for Objects in Egocentric View with a Reachability Prior

Makansi¹,

Çiçek²,

Buchicchio³

et al. 2020

Preprint

View full text Add to dashboard Cite

In this paper, we investigate the problem of anticipating future dynamics, particularly the future location of other vehicles and pedestrians, in the view of a moving vehicle. We approach two fundamental challenges: (1) the partial visibility due to the egocentric view with a single RGB camera and considerable field-of-view change due to the egomotion of the vehicle; (2) the multimodality of the distribution of future states. In contrast to many previous works, we do not assume structural knowledge from maps. We rather estimate a reachability prior for certain classes of objects from the semantic map of the present image and propagate it into the future using the planned egomotion. Experiments show that the reachability prior combined with multi-hypotheses learning improves multimodal prediction of the future location of tracked objects and, for the first time, the emergence of new objects. We also demonstrate promising zero-shot transfer to unseen datasets.

show abstract

“…By learning discriminative feature representations, deep learning has enhanced many computer vision applications such as image classification [32], video background subtraction [4], and pedestrian detection [42]. In the context of tracking, Convolutional Neural Networks (CNN) have been utilized to learn feature representations of targets instead of using heuristic and hand-crafted features [54,37,56]. CNNs have also been utilized for modeling the similarity between a pair of detections [34,52].…”

Section: Related Workmentioning

confidence: 99%

Multiple People Tracking Using Hierarchical Deep Tracklet Re-identification

Babaee,

Athar,

Rigoll

2018

Preprint

View full text Add to dashboard Cite

The task of multiple people tracking in monocular videos is challenging because of the numerous difficulties involved: occlusions, varying environments, crowded scenes, camera parameters and motion. In the tracking-by-detection paradigm, most approaches adopt person re-identification techniques based on computing the pairwise similarity between detections. However, these techniques are less effective in handling long-term occlusions. By contrast, tracklet (a sequence of detections) re-identification can improve association accuracy since tracklets offer a richer set of visual appearance and spatio-temporal cues. In this paper, we propose a tracking framework that employs a hierarchical clustering mechanism for merging tracklets. To this end, tracklet re-identification is performed by utilizing a novel multi-stage deep network that can jointly reason about the visual appearance and spatio-temporal properties of a pair of tracklets, thereby providing a robust measure of affinity. Experimental results on the challenging MOT16 and MOT17 benchmarks show that our method significantly outperforms state-of-the-arts.

show abstract

Simple Online and Realtime Tracking with a Deep Association Metric

Cited by 41 publications

References 0 publications

SoDA: Multi-Object Tracking with Soft Data Association

SoDA: Multi-Object Tracking with Soft Data Association

Multimodal Future Localization and Emergence Prediction for Objects in Egocentric View with a Reachability Prior

Multiple People Tracking Using Hierarchical Deep Tracklet Re-identification

Contact Info

Product

Resources

About