Content Based Image Retrieval From Auto Encoders Using Keras and Tensor Flow Python API A Deep Learning Technique
Content Based Image Retrieval From Auto Encoders Using Keras and Tensor Flow Python API A Deep Learning Technique
Volume 5, Issue 9, September – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Content Based Image Retrieval from Auto
Encoders Using Keras and Tensor Flow Python
API a Deep Learning Technique
Vinitha .V1, Velantina .V2
Post Graduates, University VTU, Belgaum
Abstract:- As the technology is evolving new methods The comparison of two images in content-based image
and techniques are determined and implemented in a retrieval is using an image distance measure. The image
smart way to improve and achieve a greater insight in distance measure compares the similarity of two images in
this smart era. The retrieval of image is popularly various dimensions such as color, texture, shape, and others
growing in this emerging trend. In this paper we have and the shapes is been identified by applying edge
used how to build a very simple image retrieval system detection to an image.
using a special type of Neural Network called auto
encoders. Here the images can be retrieved with visual Deep learning
contents textures, shape and this method of image Deep learning means specific subfield of machine
retrieval is called content based image retrieval. learning which is new way of learning representations from
data that puts an emphasis on model .
Keywords:- Image Retrieval, Keras, Tensor Flow, Neural
Network.
I. INTRODUCTION
Content based means the search analyzes the contents
of the image rather than the metadata such as keywords,
tags, or descriptions associated with the image.
Fig 1:- Deep Learning Layers
II. PROPOSED SYSTEM
The Auto encoders has two parts
Encoder: The network that consists of input into a latent
space representation.
Decoder: The image search engines are similar to text
search engines instead of presenting the search engine with
a text query, we can provide an image query to the image
search engine that finds all visually similar relevant images
Fig 1:- Overview of image retrieval using content based and train the auto encoder.
technique among various images.
IJISRT20SEP116 www.ijisrt.com 869
Volume 5, Issue 9, September – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Fig 3:- Structure of Content Based Image Retrieval
Train an auto encoder on the input dataset in an
unsupervised method and we use the auto encoder to extract
and store features in an index and then search the index. The
image retrieval system is build with an auto encoder that
gives importance to latent space representation vector and
the creation of an auto encoder training script is done using
Keras and Tensor Flow.
Fig 5:- The reconstructed data visualized through the auto
encoder.
Fig 4:- Training an auto encoder with Keras and Tensor
Flow for CBIR The example showing the image querying containing
the digit 9 with auto encoder image retrieval system.
IJISRT20SEP116 www.ijisrt.com 870
Volume 5, Issue 9, September – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[5]. "Different types of Autoencoders", OpenGenus IQ:
Learn Computer Science, Jul. 2019.
[6]. I. A. Siradjuddin, W. A. Wardana and M. K Sophan,
"Feature Extraction using Self-Supervised
Convolutional Autoencoder for Content based Image
Retrieval", 2019 3rd International Conference on
Informatics and Computational Sciences (ICICoS), pp.
1-5, Oct. 2019.
[7]. M. A. Nielsen, Neural Networks and Deep Learning,
[online] Available:
http://neuralnetworksanddeeplearning.com.
[8]. A. Sze-To, H. R. Tizhoosh and A. K C. Wong, "Binary
codes for tagging x-ray images via deep de-noising
autoencoders", 2016 International Joint Conference on
Neural Networks (IJCNN), pp. 2864-2871, Jul. 2016.
[9]. Keras: the Python deep learning API, [online]
Available: https://keras.io/
[10]. A. Baaj, "Keras Tutorial: Content Based Image
Retrieval Using a Convolutional Denoising
Autoencoder", Medium, Dec. 2019.
[11]. X. Yuan and C.-T. Li, "CBIR approach to building
image retrieval based on invariant characteristics in
Hough domain", 2008 IEEE International Conference
on Acoustics Speech and Signal Processing, pp.
1209-1212, Mar. 2008.
[12]. K. Ramanjaneyulu, K. V. Swamy and C. S. Rao,
"Novel CBIR System using CNN Architecture", 2018
3rd International Conference on Inventive
Fig 6:- MNIST query image. Computation Technologies (ICICT), pp. 379-383,
2018.
III. CONCLUSION
As the technology is taking a rapid swipe over the
smart world in developing and deploying the smart
technologies we have to adopt and make use of it efficiently.
We have developed model using the python libraries keras
and tensor flow. Here we are training our data and then
verifying it by providing proper inputs to which the output
will be generated. The latent space as a feature vector for the
images is calculated and the images are compared with each
other and image retrieval techniques are used in various
fileds of real world applications. Hence our model works
effectively and efficiently.
REFERENCES
[1]. N. Hubens, "Build a simple Image Retrieval System
with an Autoencoder", Medium, Aug. 2018.
[2]. M. I. Daoud, A. Saleh, I. Hababeh and R. Alazrai,
"Content-based Image Retrieval for Breast Ultrasound
Images using Convolutional Autoencoders: A
Feasibility Study", 2019 3rd International Conference
on Bio-engineering for Smart Technologies
(BioSMART), pp. 1-4, Apr. 2019.
[3]. "Autoencoders for Content-based Image Retrieval
with Keras and TensorFlow", PyImageSearch, Mar.
2020.
[4]. J. Brownlee, "Your First Deep Learning Project in
Python with Keras Step-By-Sop", Machine Leaming
Mastery, Jul. 2019.
