Realtime Vehicle and Pedestrian Tracking for Didi Udacity Self-Driving Car Challenge

Buyval, Alexander; Gabdullin, Aidar; Mustafin, Ruslan; Shimchik, Ilya

doi:10.1109/icra.2018.8460913

Cited by 24 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluated the performance of the proposed tracking method using the validation data and official evaluation tool 3 . Table II summarizes the MOTA scores obtained by the baseline and proposed method.…”

Section: A Comparison To State-of-the-art Methodsmentioning

confidence: 99%

Multi-person Pose Tracking using Sequential Monte Carlo with Probabilistic Neural Pose Predictor

Okada

Takenaka

Taniguchi

2020

2020 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

It is an effective strategy for the multi-person pose tracking task in videos to employ prediction and pose matching in a frame-by-frame manner. For this type of approach, uncertainty-aware modeling is essential because precise prediction is impossible. However, previous studies have relied on only a single prediction without incorporating uncertainty, which can cause critical tracking errors if the prediction is unreliable. This paper proposes an extension to this approach with Sequential Monte Carlo (SMC). This naturally reformulates the tracking scheme to handle multiple predictions (or hypotheses) of poses, thereby mitigating the negative effect of prediction errors. An important component of SMC, i.e., a proposal distribution, is designed as a probabilistic neural pose predictor, which can propose diverse and plausible hypotheses by incorporating epistemic uncertainty and heteroscedastic aleatoric uncertainty. In addition, a recurrent architecture is introduced to our neural modeling to utilize time-sequence information of poses to manage difficult situations, such as the frequent disappearance and reappearances of poses. Compared to existing baselines, the proposed method achieves a state-of-the-art MOTA score on the PoseTrack2018 validation dataset by reducing approximately 50% of tracking errors from a state-of-the art baseline method.

show abstract

Section: A Comparison To State-of-the-art Methodsmentioning

confidence: 99%

Multi-person Pose Tracking using Sequential Monte Carlo with Probabilistic Neural Pose Predictor

Okada

Takenaka

Taniguchi

2020

2020 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

show abstract

“…The results for the Control scenario show that the maximum response time (MAXRT)-in the 12 executions-is 44 ms. This is considered as near to real-time response time 61 (average latency for LTE-Advance is 45 ms 62 ). The average response time for successful requests (AVGRS) shows similar results in all the scenarios, ranging from an average of 18 ms in the Control scenario, to 27 ms reached in both the KillPods and DrainNodes scenarios.…”

Section: Guestbook User Facing Applicationmentioning

confidence: 99%

Chaos as a Software Product Line—A platform for improving open hybrid‐cloud systems resiliency

et al. 2022

View full text Add to dashboard Cite

Nowadays, cloud-native software architectures have a significant relevance due to the speed and agility they provide. These properties lead relevant organizations in different industries, like video streaming (Netflix), car-sharing (Uber, Cabify), banking (BBVA, HSBC), and governmental agencies (NASA, FBI, CERN, ESA) to heavily rely on cloud-native software to run their business-critical applications. Additionally, including fault injection actions in the production infrastructure allows companies to have consistent environments, to improve applications dependability against unexpected failures, to provide better user experience, and to improve the overall system quality. Thus, cloud computing technologies allow development teams to rapidly create complex systems and to continuously deploy them, at a global scale. This work describes Pystol, a novel fault injection platform-represented as aSoftware Product Line-to analyze the effects caused by a wide spectrum of adverse conditions. Pystol is designed to be executed on top of cloud-native environments, either in private or public clouds. The proposed architecture shows a way for representing feature models based on Unified Model Language (in short, UML) component diagrams. Furthermore, we present a thorough empirical study carried out in real-world environments, providing promising results.

show abstract

“…The failure rate for two-stream CNN is low and thus the track life span is longer which has resulted in slightly lower average accuracy. We also would to point out that most of the tracking applications such as physiotherapy analysis [33] and self-driving car [34] emphasis more on the tracking robustness especially for the case of short term tracking. The aforementioned systems need to still functioning well albeit lower accuracy, rather than having a tracking failure which will totally affect the systems accuracy.…”

Section: Benchmark Validationmentioning

confidence: 99%

Two Streams Multiple-Model Object Tracker for Thermal Infrared Video

Zulkifley

2019

IEEE Access

View full text Add to dashboard Cite

Thermal infrared (TIR) visual object tracking has been applied in various applications, such as pedestrian detection, wildlife observation, surveillance systems, and so on. The tracker function is to track a particular object of interest and generate its trajectory which will be integrated into a decision making process. Many trackers that are based on fully convolutional neural networks (CNNs) have good performance for RGB input, but it is not the case for the TIR input. Its lack of texture information and the fact that it produces similar heat maps between two nearby objects make the tracking task very challenging. By relying on a fully CNN network alone, a tracker can learn the appearance model but it will not work well if the object heat map looks too similar to the background. Hence, a Siamese CNN network can be implemented to complement the fully CNN, as it allows a set of recent object templates to be used for matching purposes. Yet, the Siamese network alone is not accurate especially in the case of occlusions, as the stored templates rarely produce robust matching. Thus, we propose a two-stream CNN tracker that combines the fully CNN and the Siamese CNN such that each network keeps a set of matching models to cater to diverse appearance changes. Furthermore, the CNN layers are shared between both CNN streams to reduce computational burden. A single dense score map is produced by overlaying the normalized scores of the two streams. The experiments on VOT-TIR 2016 database show that our tracker works well for the datasets with high motion blur, occlusion, and appearance deformation. Besides, the Siamese CNN response map can also be used as an indicator to decide the size of the search region. INDEX TERMS Siamese networks, single object tracking, thermal infrared video.

show abstract

Realtime Vehicle and Pedestrian Tracking for Didi Udacity Self-Driving Car Challenge

Cited by 24 publications

References 9 publications

Multi-person Pose Tracking using Sequential Monte Carlo with Probabilistic Neural Pose Predictor

Multi-person Pose Tracking using Sequential Monte Carlo with Probabilistic Neural Pose Predictor

Chaos as a Software Product Line—A platform for improving open hybrid‐cloud systems resiliency

Two Streams Multiple-Model Object Tracker for Thermal Infrared Video

Contact Info

Product

Resources

About