Attendance Marking System Using Image Recognition: Professor: Sanjay Srivastava

ATTENDANCE MARKING SYSTEM USING
IMAGE RECOGNITION
Professor : Sanjay Srivastava

Team :
Kunal Suthar(201401131)
Nirav Patel(201401179)
Maulik Limbadiya(201401189)
Introduction Technology used
● OpenCV : An open-source library
Our proposed solution attempts to mark
which allows us to use the HAAR
attendance of students during lecture
cascade which is used for face
hours by using multiple cameras at multiple
detection and pre-processing the
places, which would take photos of the
image.
assigned portion of the class at multiple
● Keras : An open source neural
time intervals. These photos are stored at a
network based library used for image
server, where the face detection and
recognition.
recognition process commences.
● Python 2.7
Implementation Steps
● Creation of database and training the convolutional neural network.

● Image acquisition from real environment in the form of frames.
● Face detection from the frames.
● Face recognition and testing of the neural net.
● Finding adequate image size and quality for achieving better detection
and recognition accuracy.
Face Detection
Object Detection using Haar feature-based cascade classifiers is a machine learning based approach
where a cascade function is trained from a lot of positive and negative images. Then we need to extract
features from it.
Four basic types of features:
– They are easy to calculate.

– The white areas are subtracted from the black ones.
– A special representation of the sample called the integral image makes feature extraction faster.
Integral Images
● Summed area tables
● A representation that means any rectangle’s area can be calculated in four indexes to the integral image
Feature Extraction
1. Features are extracted from sub windows of an sample image.

–The base size for a sub window is 24 by 24 pixels.
•In a 24 pixel by 24 pixel sub window there are 180,000 possible features to be
calculated.
2. What is the end result of feature extraction?

–A lot of data!
•This is called overfitting and the amount of data must be reduced.
•Overfitting can be compensated to an extent by logical elimination.
Feature/Classifier Evaluation
● Using AdaBoost the number of features is dramatically reduced.
–A simple algorithm that selects one feature at a time and assigns weights to the
feature. It helps in combining multiple “weak classifiers” and thus producing a strong
classifier.
● Thus, these features helps us in face detection.

Face Recognition
We can train a neural network to identify images. But the problem faced in
the image recognition is that the image is not always centered, as we had in
our training data. We can use the concept of a sliding window, but it’s
inefficient. Instead, we introduce more data where the face can be
anywhere in the image. We add more layers in the neural network. But, this
is also inefficient. So, instead of training neural network with more number
of data tuple, we increase the network size itself to reduce computation.
Working of convolution
Similar to a sliding window, we pass a window through an image and save the result of each
window as a separate image. Feed each image in the small neural network. Image which
gives the best classification results are selected. Array size is reduced. To reduce it further,
we do max pooling. We look at the 2*2 window and the image with the highest weight
among these are selected. And finally, it is applied to neural network.
Results
Face Detection : For images of good quality, the face detection algorithm provides pretty
high accuracy. Following are some prerequisite constraints :
● The size of faces in images should be of at least 25*25 resolution.

● Blurred faces do get detected but cannot recognized. So blurred images are useless.
● The images shouldn’t be too dark.
● Side faces aren’t detected.
The results of face detection algorithm that we ran on a video recorded in a regular class
are shown below. The given results are for a 30 fps video from which a frame from every 5
frames is extracted for face detection.
Following are two examples where some students who are not detected in first image get
detected in the second image. So, a stream of such images covers almost all of the students.
Face Recognition Results:
Following are the results for face recognition which we ran on an image database that we
downloaded from internet particularly for face recognition. For each person around 30
images were available to train the model. Each image had a little variation like angle,
sunglasses, facial expression. And because the quality of the images of given database was
good and the total number of distinct persons were less, the accuracy is pretty high for only
10 epochs. For a fairly large amount of class, the total number of distinct student increases
and so the number epochs to train the model increases too. This also gives an idea about
the data collection process for the given system.
Conclusion/Future Extension
● What we did is just the groundwork for implementing the attendance system i.e.,
implementing the software part of it.
● A network of cameras and server is to be setup to actually implement the system.
● Deployability in an actual class is yet to be tested.
References
1. An Automatic Attendance System Using Image processing by Aziza Ahmedi, Dr Suvarna

Nandyal in The International Journal Of Engineering And Science (IJES).
2. Real-Time Face Detection Using Gentle AdaBoost Algorithm and Nesting Cascade
Structure by Jian-qing Zhu and Can-hui Cai at 2012 IEEE International Symposium on
Intelligent Signal Processing and Communication Systems (ISPACS 2012) November 4-7,
2012.
3. Rapid Object Detection using a Boosted Cascade of Simple Features by Paul Viola and
Michael Jones at 2001 IEEE.
4. Real Time Face Recognition Using Adaboost Improved Fast PCA Algorithm. International
Journal of Artificial Intelligence & Applications (IJAIA), Vol.2, No.3, July 2011
5. Medium Articles for the convolutional neural network part:
https://medium.com/@ageitgey/machine-learning-is-fun-part-3-deep-learning-and-convol
utional-neural-networks-f40359318721

Attendance Marking System Using Image Recognition: Professor: Sanjay Srivastava

Uploaded by

Copyright:

Available Formats

Attendance Marking System Using Image Recognition: Professor: Sanjay Srivastava

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Attendance Marking System Using Image Recognition: Professor: Sanjay Srivastava

Uploaded by

Copyright:

Available Formats

ATTENDANCE MARKING SYSTEM USING

Professor : Sanjay Srivastava

● Creation of database and training the convolutional neural network.

Four basic types of features:

– They are easy to calculate.

1. Features are extracted from sub windows of an sample image.

2. What is the end result of feature extraction?

● Thus, these features helps us in face detection.

● The size of faces in images should be of at least 25*25 resolution.

1. An Automatic Attendance System Using Image processing by Aziza Ahmedi, Dr Suvarna

You might also like