Scribd
Upload
30 views9 pages

Synopsis

The document discusses a project synopsis on pedestrian detection. It provides an overview of handcrafted and deep features for pedestrian identification. It discusses the methodology, literature review, problem formulation and objectives, hardware and software requirements, and references.

Uploaded by

mohammedsaifulla010
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
30 views9 pages

Synopsis

The document discusses a project synopsis on pedestrian detection. It provides an overview of handcrafted and deep features for pedestrian identification. It discusses the methodology, literature review, problem formulation and objectives, hardware and software requirements, and references.

Uploaded by

mohammedsaifulla010
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 9

ST.

PHILOMENA’S COLLEGE (AUTONOMOUS)


College of Excellence Bannimantap, Mysore-570015

A Project Synopsis
An Overwiew of Deep and Handcrafted Features with Redestrain Identification
Submitted as partial fulfilment of the requirement for the award of the Degree in

BACHELORS OF COMPUTER APPLICATION

By

Musadiq Shariff
SB210365

Final year

Dept. of Computer Science

St. Philomena’s College (Autonomous) College of Excellence


Bannimantap, Mysore - 570015
Abstract—Pedestrian detection is an important but challenging problem in
computer vision, especially in human-centric tasks. Over the past decade,
significant improvement has been witnessed with the help of handcrafted
features and deep features. Here we present a comprehensive survey on recent
advances in pedestrian detection. First, we provide a detailed review of single-
spectral pedestrian detection that includes handcrafted features based methods
and deep features based approaches. For handcrafted features based methods,
we present an extensive review of approaches and find that handcrafted
features with large freedom degrees in shape and space have better
performance. In the case of deep features based approaches, we split them into
pure CNN based methods and those employing both handcrafted and CNN
based features. We give the statistical analysis and tendency of these methods,
where feature enhanced, part-aware, and post-processing methods have
attracted main attention. In addition to single-spectral pedestrian detection, we
also review multi-spectral pedestrian detection, which provides more robust
features for illumination variance. Furthermore, we introduce some related
datasets and evaluation metrics, and a deep experimental analysis. We
conclude this survey by emphasizing open problems that need to be addressed
and highlighting various future directions.

Methodology/ Planning of work:

(a) Proposal generation: This step aims to extract some candidate proposals of
pedestrians from an input image. The proposals indicate a set of bounding-boxes
which po- tentially represent the objects. Common strategies include sliding-
window methods , particle-window methods , objectness methods , and region
proposal network, The sliding-window methods (SW) adopt a greedy search strat-
egy with a fixed-sized step to scan the image from the top- left to bottom-right
region. The particle-window methods adopt the coarse-to-fine cascaded search
where the propos- als generated at current stage follow the likelihood distri-
bution of previous stage. The objectness methods typically employ a variety of
low-level features (e.g., edge and color features) to extract the proposals in a
bottom-up fashion. Recently, a region proposal network (RPN) was introduced for
proposal generation, which shares the deep features with the following proposal
classification and regression.
(b) Proposal classification: This step assigns these candi- date proposals to the
positive class (pedestrians) or the negative class (background) based on the
extracted features of these proposals. The handcrafted features based methods
adopt a shallow classifier (e.g., SVM or boosting) for classification, whereas
deep features based methods generally integrate the feature extraction and
classification into a unified framework by utilizing a soft- max (or sigmoid) layer.
Additionally, deep features based methods add regression in parallel with
classification to refine the location quality of the bounding-boxes.
(c) Post processing: As a single pedestrian may be detected by multiple
bounding-boxes after proposal
(d) classification, which is the issue of duplicate detections. The technique of
non-maximum suppression (NMS) selects the best bounding-box for each object
and suppresses other duplicate bounding-boxes. The related methods can be di-
vided into two categories: heuristic-based and learning- based methods. The
heuristic-based methods (e.g., greedy NMS, Soft-NMS , SGE-NMS , and
Adaptive NMS) combine the bounding-boxes according to classifi- cation scores,
where the overlapped bounding-boxes with lower scores are suppressed. The
learning-based methods, including Gnet and Relation Network , learn a mapping
to retain the most accurate bounding-boxes.

Literature review:

Title of The paper Description Publication details

Jeonghyun Baek, When pedestrian detection IEEE Transactions on


Junhyuk Hyun, and (PD) is implemented on a Intelligent Transportation
Euntai Kim. A central processing unit Systems, 21(3):1216–1228,
pedestrian (CPU), performing real-time 2020
detection system processing using a classical
accelerated by sliding window is difficult.
kernelized Therefore, an efficient
proposals. proposal generation method is
required. A new generation
method, named additive
kernel binarized normed
gradient (AKBING), is
proposed herein, and this
method is applied to the PD
for real-time implementation
on a CPU.

Xuangeng Chu, We propose a simple yet Proc. IEEE Conference on


Anlin Zheng, effective proposal-based Computer Vision and Pattern
Xiangyu Zhang, object detector, aiming at Recognition, 2020
and Jian Sun. detecting highly-overlapped
Detection in instances in crowded scenes.
crowded scenes: The key of our approach is to
One proposal, let each proposal predict a set
multiple of correlated instances rather
predictions. than a single one in previous
proposal-based frameworks.

My Kieu, Andrew Pedestrian detection is a core Proc. European Conference on


D. Bagdanov, problem in computer vision, Computer Vision, 2020
Marco Bertini, and and is a problem that is
Alberto del Bimbo. gaining prominence due to its
Task-conditioned importance in assisted and
domain adaptation autonomous driving
for pedestrian applications. Many state-of-
detection in the-art approaches, especially
thermal imagery. those used for autonomous
driving, combine thermal and
visible spectrum imagery in
order to robustly detect
persons independent of time
of day or weather conditions.

Problem formulation/Objectives:

Most vision applications, including pedestrian detection, acquire data using


visible-light cameras since they are inex- pensive and easily available. As such,
most existing pedes- trian detection methods employ this kind of data. We further
separate these pedestrian detection methods into two main categories:
handcrafted features based approaches and deep features based ap- proaches.
Moreover, the deep features based approaches are split into pure CNN based
methods and hybrid methods. Next, we discuss the handcrafted features methods
and then present a summary of deep features based methods.

In past decade, pedestrian detection has witnessed signifi- cant success, which has
gone from the handcrafted features to deep features based approaches. In this
paper, we first summarize these two types of methods in detail. After- wards, we
review multispectral pedestrian detection. We review popular pedestrian datasets
and a deep analysis on pedestrian detection methods. Finally, we discuss some
challenging problems (e.g., occlusion, scale variance, and domain adaptation) in
pedestrian detection. We hope that this deep survey can help the researchers to
develop new methods in the field of pedestrian detection.

HARDWARE REQUIREMENTS:

 System : I3 and above

 Hard Disk : 500GB.

 Monitor : 15 inch VGA Color.

 RAM : 4GB

SOFTWARE REQUIREMENTS:
 Operating System : Windows 10 and above.

 Programming Language : Python 3

 Tools :Anaconda,Spyder

References and bibliography:

 Anelia Angelova, Alex Krizhevsky, Vincent Vanhoucke1, Abhijit Ogale,


and Dave Ferguson. Real-time pedestrian detection with deep network
cascades. Proc. British Machine Vision Conference, 2015.Jeonghyun Baek,
Junhyuk Hyun, and Euntai Kim.

 A pedestriandetection system accelerated by kernelized proposals. IEEE


Transactions on Intelligent Transportation Systems, 21(3):1216–1228,
2020.Herbert Bay, Tinne Tuytelaars, and Luc Van Gool. Surf: Speededup
robust features. Proc. European Conference on Computer Vision,
2006.Rodrigo Benenson, Markus Mathias, Radu Timofte, and Luc Van

 Pedestrian detection at 100 frames per second. Proc. IEEE Conference on


Computer Vision and Pattern Recognition, 2012.Rodrigo Benenson,
Markus Mathias, Tinne Tuytelaars, and Luc Van Gool. Seeking the
strongest rigid detector. Proc. IEEE Conference on Computer Vision and
Pattern Recognition, 2013.Rodrigo Benenson, Mohamed Omran, Jan
Hosang, and Bernt Schiele.

 Ten years of pedestrian detection, what have we learned? Proc. European


Conference on Computer Vision, 2014. [7]Navaneeth Bodla, Bharat Singh,
Rama Chellappa, and Larry S.Davis. Soft-nms–improving object detection
with one line of code. Proc. IEEE International Conference on Computer
Vision, 2017.Markus Braun, Sebastian Krebs, Fabian Flohr, and Dariu M.
Gavrila. Eurocity persons:

 A novel benchmark for person de- tection in traffic scenes. IEEE


Transactions on Pattern Analysis and Machine Intelligence, 41(8):1844–
1861, 2019.Garrick Brazil and Xiaoming Liu.

 Pedestrian detection with autoregressive network phases. Proc. IEEE Conference


on Computer Vision and Pattern Recognition, 2019. Garrick Brazil, Xi Yin, and
Xiaoming Liu. Illuminating pedes-trians via simultaneous detection and
segmentation. Proc. IEEE International Conference on Computer Vision, 2017.

You might also like