Improving Reliability of Object Detection for Lunar Craters Using Monte Carlo Dropout

Myojin, Tomoyuki; Hashimoto, Shintaro; Mori, Kenji; Sugawara, Keisuke; Ishihama, Naoki

doi:10.1007/978-3-030-30508-6_6

Cited by 11 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similarly, He et al proposed a new Kullback-Leibler (KL) loss to learn localization (for example, variance) of the uncertainty with border boxes, which can enable a voting plan to select border boxes [33]. Another work is to run YOLOv3 using MC drop on a large scale towards the dataset of the pedestrian using the variance defined by MC drop in YOLOv3 architecture to measure the spatial uncertainty [34]. As a result, recognition of cavities can be accepted or rejected based on that variance.…”

Section: Uncertainty Of Object Recognitionmentioning

confidence: 99%

Effective Parameters to Design an Automatic Parking System

Razavi

Sherafati

2022

UJCSC

View full text Add to dashboard Cite

The automated parking system is an extensive branch of smart transport systems. The smartness of such systems is determined by different parameters such as parking maneuver planning. Coding this control system includes vehicle parking and understanding the environment. A high-quality classification mask has been used on each sample to analyze the automated vehicle parking parameters. Mask region-based convolutional neural networks (R-CNN) was taught using a small computational workload titled faster R-CNN that operates in five frames per second. In this paper, the rapidly-exploring random tree (RRT) method was used for routing the parking space and a nonlinear model predictive control (NMPC) controller was added to develop this system. We add the line detection algorithm commands to the mask R-CNN algorithm. The results can be useful to design a secure automatic parking system as well as a powerful perception system.

show abstract

Section: Uncertainty Of Object Recognitionmentioning

confidence: 99%

Effective Parameters to Design an Automatic Parking System

Razavi

Sherafati

2022

UJCSC

View full text Add to dashboard Cite

show abstract

“…Other techniques such as the temperature scaling technique can be applied to calibrate the confidence on final outputs [8]. The Monte-Carlo dropout technique [9], [10] is an online method that utilizes dropout to create ensembles and to compute the Bayesian measure. The above methods are essentially confidence-based and require proper calibration in order to be used in safety-critical systems 1 .…”

Section: Related Workmentioning

confidence: 99%

Monitoring Object Detection Abnormalities via Data-Label and Post-Algorithm Abstractions

Chen¹,

Cheng²,

Yan³

et al. 2021

Preprint

View full text Add to dashboard Cite

While object detection modules are essential functionalities for any autonomous vehicle, the performance of such modules that are implemented using deep neural networks can be, in many cases, unreliable. In this paper, we develop abstraction-based monitoring as a logical framework for filtering potentially erroneous detection results. Concretely, we consider two types of abstraction, namely data-label abstraction and post-algorithm abstraction. Operated on the training dataset, the construction of data-label abstraction iterates each input, aggregates region-wise information over its associated labels, and stores the vector under a finite history length. Post-algorithm abstraction builds an abstract transformer for the tracking algorithm. Elements being associated together by the abstract transformer can be checked against consistency over their original values. We have implemented the overall framework to a research prototype and validated it using publicly available object detection datasets.

show abstract

“…Another work [8] on a large scale automotive pedestrian dataset also reimplemented YOLOv3 in order to estimate epistemic and aleatoric uncertainty by using MC-Drop. Finally, another related work [17] used the variance introduced by MC-Drop on a YOLOv3 architecture to measure spatial uncertainty; as a result, detections of lunar craters could be accepted or rejected on the basis of that variance. However, the spatial quality of those bounding boxes were not directly evaluated using specific quantitative metrics and these detections did not need to be performed in real-time nor in safety-critical situations like Stochastic-YOLO is targeted for.…”

Section: Related Workmentioning

confidence: 99%

Stochastic-YOLO: Efficient Probabilistic Object Detection under Dataset Shifts

Azevedo,

de Jong,

Mattina

et al. 2020

Preprint

View full text Add to dashboard Cite

In image classification tasks, the evaluation of models' robustness to increased dataset shifts with a probabilistic framework is very well studied. However, Object Detection (OD) tasks pose other challenges for uncertainty estimation and evaluation. For example, one needs to evaluate both the quality of the label uncertainty (i.e., what?) and spatial uncertainty (i.e., where?) for a given bounding box, but that evaluation cannot be performed with more traditional average precision metrics (e.g., mAP). In this paper, we adapt the well-established YOLOv3 architecture to generate uncertainty estimations by introducing stochasticity in the form of Monte Carlo Dropout (MC-Drop), and evaluate it across different levels of dataset shift. We call this novel architecture Stochastic-YOLO, and provide an efficient implementation to effectively reduce the burden of the MC-Drop sampling mechanism at inference time. Finally, we provide some sensitivity analyses, while arguing that Stochastic-YOLO is a sound approach that improves different components of uncertainty estimations, in particular spatial uncertainties.

show abstract

Improving Reliability of Object Detection for Lunar Craters Using Monte Carlo Dropout

Cited by 11 publications

References 7 publications

Effective Parameters to Design an Automatic Parking System

Effective Parameters to Design an Automatic Parking System

Monitoring Object Detection Abnormalities via Data-Label and Post-Algorithm Abstractions

Stochastic-YOLO: Efficient Probabilistic Object Detection under Dataset Shifts

Contact Info

Product

Resources

About