What is the important aspect of the multi-object detection system?

Relative locations of the objects provide constraints that help to make the system more robust by focusing the search in regions where the object is expected based on locations of the other objects.

How many samples are obtained with the second strategy?

In the second strategy, the authors obtain up to M = 2000 samples with the strongest weight and perform k-means clustering to get N number of modes.

What is the difficult part of the multi-object detection task?

Since obtaining a joint model of multiple objects is difficult in most practical situations, the multi-object detection task has been solved by multiple individual object detectors connected by a spatial model [4].

What is the likelihood of a hypothesized state that gives rise to observations?

The likelihood of a hypothesized state that gives rise to observations is based on a deterministic model learned using a large annotated database of images.

How do the authors estimate the probability of a state?

At each time step t, the estimation involves sampling from the proposal distribution p(xt|x0:t−1,y0:t) of the current state xt conditioned on the history of states x0:t−1 up to time t − 1 and the history of observations y0:t up to time t.The authors also use sequential Monte Carlo technique in multiobject detection.

What is the way to model f(t|0:t1)?

In this paper, the authors use a pairwise dependencyf(θt|θ0:t−1) = f(θt|θj), j ∈ {0, 1, . . . , t− 1}. (13)The authors model f(θt|θ0:t−1) as a Gaussian distribution estimated from the training data.

What is the likelihood of a hypothesized state?

The computation requires a likelihood of a hypothesized state that gives rise to observations and a transition model that describes the way states are propagated between frames.

How is the multi-object detection problem solved?

The multi-object detection problem is solved by recursively applying prediction and update steps to obtain the posterior distribution f(θ0:t|V0:t).

Why is the order of the detection not preferred?

Since in real detection scenarios the ground truth is not available and sampling in low-probability regions is not reliable, these sequential orders are not preferred.

What is the order of detection of a system?

The order of detection is automatically determined by a greedy algorithm that puts the most reliable detections earlier in the detection sequence.

What are the important questions for the multiscale scheduling?

The important questions are how to determine the size of the context region (detection scale) and which objects to detect first in an optimal way.

What is the performance of the multi-object detection system?

The performance of their multi-object detection system is further improved by starting from objects that are easier to detect1and constraining the detection of the other objects by exploiting object configurations.

Open AccessProceedings Article10.1109/CVPR.2010.5539842

Multiple object detection by sequential monte carlo and Hierarchical Detection Network

Michal Sofka, +3 more

- 13 Jun 2010

- pp 1735-1742

TL;DR: This paper presents an algorithm that optimally selects the order based on probability of states (object poses) within the ground truth region, and shows on 2D ultrasound images of left atrium, that the automatically selected sequential order yields low mean detection error.

Abstract: In this paper, we propose a novel framework for detecting multiple objects in 2D and 3D images. Since a joint multi-object model is difficult to obtain in most practical situations, we focus here on detecting the objects sequentially, one-by-one. The interdependence of object poses and strong prior information embedded in our domain of medical images results in better performance than detecting the objects individually. Our approach is based on Sequential Estimation techniques, frequently applied to visual tracking. Unlike in tracking, where the sequential order is naturally determined by the time sequence, the order of detection of multiple objects must be selected, leading to a Hierarchical Detection Network (HDN). We present an algorithm that optimally selects the order based on probability of states (object poses) within the ground truth region. The posterior distribution of the object pose is approximated at each step by sequential Monte Carlo. The samples are propagated within the sequence across multiple objects and hierarchical levels. We show on 2D ultrasound images of left atrium, that the automatically selected sequential order yields low mean detection error. We also quantitatively evaluate the hierarchical detection of fetal faces and three fetal brain structures in 3D ultrasound images.

AI Agents for this Paper

Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps

Most frequently asked questions

1. What contributions have the authors mentioned in the paper "Multiple object detection by sequential monte carlo and hierarchical detection network" ?

In this paper, the authors propose a novel framework for detecting multiple objects in 2D and 3D images.. Their approach is based on Sequential Estimation techniques, frequently applied to visual tracking.. The authors present an algorithm that optimally selects the order based on probability of states ( object poses ) within the ground truth region.. The authors show on 2D ultrasound images of left atrium, that the automatically selected sequential order yields low mean detection error.

2. What have the authors stated for future works in "Multiple object detection by sequential monte carlo and hierarchical detection network" ?

The described framework opens up several possible avenues of future research.. This will result in a stronger geometrical constraint and therefore improve performance on objects that are difficult to detect by exploiting only the pairwise dependence.

3. What is the way to detect a large object?

During detection, larger object context is considered at coarser image resolutions resulting in robustness against noise, occlusions, and missing data.

4. What is the method for estimating the likelihood of a state?

Since the likelihood models in practical situations lead to intractable inference, approximation by Monte Carlo methods, also known as particle filtering, have been widely adopted.

Figure 1. Examples of multi-object detection: five landmarks of left atrium (LA) apical two chamber (A2C) view (left) and 3D ultrasound volume of fetal brain with three anatomies (right).

Figure 8. The detection order and the hierarchy of three brain structures: Cerebellum (CER), Cisterna Magna (CM), and Lateral Ventricles (LV). Scale selection is applied.

Table 1. Measurement errors of the hierarchical detection system (top part of the table) compared to an earlier system without the hierarchy [2]. Mean error, standard deviation, median error, and maximum error are computed. The system was trained using number of volumes specified in the 6th column and tested on the number of volumes specified in the 7th column. The average detection error using the hierarchy is 2.2 mm on data with 1 mm finest resolution. The average error of the system without the hierarchy is 4.8 mm.

Figure 7. Comparing automatically selected order (red) to 100 randomly selected orders. The final detection errors were averaged over all testing images and detected landmarks. The score indicates preference of a particular order. The automatically selected order has a low mean detection error and a high score.

Figure 9. Example results of the fetal face detection using a hierarchy of three resolutions. Initial pose after loading the volume (top row), after automatic detection at the finest level (middle row), and after volume carving of the region in front of the face (bottom row).

Figure 2. In multi-object detection, the set of observations is a sequence of image patches. The sequence specifies a spatial order rather than a time order. The latter is typically exploited in tracking applications.

Citations

Book Chapter•10.1007/978-3-642-38868-2_38

Rapid multi-organ segmentation using context integration and discriminative models

Nathan Lay, +3 more

- 28 Jun 2013

TL;DR: A novel framework for rapid and accurate segmentation of a cohort of organs that integrates local and global image context through a product rule to simultaneously detect multiple landmarks on the target organs and exploits sparsity in the global context for efficient detection.

...read moreread less

•Journal Article•10.1016/J.MEDIA.2019.04.002

Computational anatomy for multi-organ analysis in medical imaging: A review.

Juan J. Cerrolaza, +6 more

- 01 Aug 2019

- Medical Image Analysis

TL;DR: A review of the state of the art on multi-organ analysis and associated computation anatomy methodology is presented in this article, with a methodology-based classification of different techniques available for the analysis of multi-organs and multi-anatomical structures, from techniques using point distribution models to the most recent deep learning-based approaches.

...read moreread less

•Journal Article•10.1109/TBME.2012.2237400

Automatic Segmentation of Antenatal 3-D Ultrasound Images

Jérémie Anquez, +3 more

- 01 Jan 2013

- IEEE Transactions on Biomedical Engineer...

TL;DR: The segmentation method was shown to be robust to different types of initialization and to provide accurate results, with an average overlap measure of 0.89 when comparing with manual segmentations.

...read moreread less

•Journal Article•10.1109/TMI.2014.2301936

Automatic Detection and Measurement of Structures in Fetal Head Ultrasound Volumes Using Sequential Estimation and Integrated Detection Network (IDN)

Michal Sofka, +4 more

- 21 Jan 2014

- IEEE Transactions on Medical Imaging

TL;DR: An automatic fetal head and brain (AFHB) system for automatically measuring anatomical structures from 3-D ultrasound volumes is proposed, based on a novel framework for detecting multiple structures in3-D volumes.

...read moreread less

Patent

Data Transmission in Remote Computer Assisted Detection

Michal Sofka, +5 more

- 06 Apr 2011

TL;DR: In this paper, a hierarchal detection is used, allowing detection on data at progressively greater resolutions, by limiting the number and/or size of regions provided at higher resolutions based on the previous detection.

...read moreread less

...

Expand

References

•Proceedings Article•10.1109/CVPR.2005.177

Histograms of oriented gradients for human detection

Navneet Dalal, +1 more

- 20 Jun 2005

TL;DR: It is shown experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection, and the influence of each stage of the computation on performance is studied.

...read moreread less

36.7K

Journal Article•10.1023/B:VISI.0000013087.49260.FB

Robust Real-Time Face Detection

Paul A. Viola, +1 more

- 01 May 2004

- International Journal of Computer Vision

TL;DR: In this paper, a face detection framework that is capable of processing images extremely rapidly while achieving high detection rates is described. But the detection performance is limited to 15 frames per second.

...read moreread less

14.6K

Proceedings Article•10.1109/ICCV.2001.937709

Robust real-time face detection

Paul A. Viola, +1 more

- 07 Jul 2001

TL;DR: A new image representation called the “Integral Image” is introduced which allows the features used by the detector to be computed very quickly and a method for combining classifiers in a “cascade” which allows background regions of the image to be quickly discarded while spending more computation on promising face-like regions.

...read moreread less

13.8K

•Journal Article•10.1109/TPAMI.2009.167

Object Detection with Discriminatively Trained Part-Based Models

Pedro F. Felzenszwalb, +3 more

- 01 Sep 2010

- IEEE Transactions on Pattern Analysis an...

TL;DR: An object detection system based on mixtures of multiscale deformable part models that is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges is described.

...read moreread less

11.9K

Book•10.1007/978-1-4757-3437-9

Sequential Monte Carlo methods in practice

Arnaud Doucet, +3 more

- 01 Jan 2001

TL;DR: This book presents the first comprehensive treatment of Monte Carlo techniques, including convergence results and applications to tracking, guidance, automated target recognition, aircraft navigation, robot navigation, econometrics, financial modeling, neural networks, optimal control, optimal filtering, communications, reinforcement learning, signal enhancement, model averaging and selection.

...read moreread less

7.6K

...

Expand

Multiple object detection by sequential monte carlo and Hierarchical Detection Network

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What contributions have the authors mentioned in the paper "Multiple object detection by sequential monte carlo and hierarchical detection network" ?

2. What have the authors stated for future works in "Multiple object detection by sequential monte carlo and hierarchical detection network" ?

3. What is the way to detect a large object?

4. What is the method for estimating the likelihood of a state?

5. What is the important aspect of the multi-object detection system?

6. How many samples are obtained with the second strategy?

7. What is the difficult part of the multi-object detection task?

8. What is the likelihood of a hypothesized state that gives rise to observations?

9. How do the authors estimate the probability of a state?

10. What is the way to model f(t|0:t1)?

11. What is the likelihood of a hypothesized state?

12. How is the multi-object detection problem solved?

13. Why is the order of the detection not preferred?

14. What is the order of detection of a system?

15. What are the important questions for the multiscale scheduling?

16. What is the performance of the multi-object detection system?

Figures

Citations

Rapid multi-organ segmentation using context integration and discriminative models

Computational anatomy for multi-organ analysis in medical imaging: A review.

Automatic Segmentation of Antenatal 3-D Ultrasound Images

Automatic Detection and Measurement of Structures in Fetal Head Ultrasound Volumes Using Sequential Estimation and Integrated Detection Network (IDN)

Data Transmission in Remote Computer Assisted Detection

References

Histograms of oriented gradients for human detection

Robust Real-Time Face Detection

Robust real-time face detection

Object Detection with Discriminatively Trained Part-Based Models

Sequential Monte Carlo methods in practice

Related Papers (5)

Probabilistic boosting-tree: learning discriminative models for classification, recognition, and clustering

Hierarchical, learning-based automatic liver segmentation

Four-Chamber Heart Modeling and Automatic Segmentation for 3-D Cardiac CT Volumes Using Marginal Space Learning and Steerable Features

Active shape models—their training and application

Regression forests for efficient anatomy detection and localization in computed tomography scans