ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
Dog Breed Classification Using Convolutional
Neural Network
Suyash B1, Rishikesh P2, Rohit W3, Kaustubh J4, Prof. Balaji Bodhke5
Students, Department of Computer Engineering1,2,3,4
Faculty, Department of Computer Engineering5
Modern Education Society College of Engineering, Pune, India
Abstract: Dogs are one of the most common domestic animals. Due to a large number of dogs, there are
several issues such as population control, decreased outbreak such as Rabies, vaccination control, and
legal ownership. At present, there are over 180 dog breeds. Each dog breed has specific characteristics
and health conditions. In order to provide appropriate treatments and training, it is essential to identify
individuals and their breeds. Machine learning gives the strength on the way to train algorithms model
that can handle the difficulties of info classification also prediction grounded on totally on arising
information as of raw information. Convolutional Neural Networks ( CNNs ) gives single often used
methods for image classification and detection. In this exertion , we define a CNN based approach
for spotting dogs in perchance complex images and due to this fact reflect inconsideration on the
identification of the one of kinds of dog breed. The experimental outcome analysis supported the
standard metrics and thus the graphical representation confirms that the algorithm ( CNN ) gives good
analysis accuracy for all the tested datasets.
Keywords: CNN, Machine Learning, Classification
I. INTRODUCTION
Machine learning is a subfield of artificial intelligence (AI). The goal of machine learning generally is to understand
the structure of data and fit that data into models that can be understood and utilized by people. Although machine
learning is a field within computer science, it differs from traditional computational approaches. In traditional
computing, algorithms are sets of explicitly programmed instructions used by computers to calculate or problem solve.
Machine learning algorithms instead allow for computers to train on data inputs and use statistical analysis in order to
output values that fall within a specific range. Because of this, machine learning facilitates computers in building
models from sample data in order to automate decision-making processes based on data inputs. Most of the dog breeds
are developed in order to drive some specific things. Knowing the breed of dog can help us to predict and understand
the behavior. And this is essential when it comes to managing and training dogs for specific tasks. In machine learning,
convolutional Neural Network (CNN) is complicated feed forward neural networks. CNNs are used for image
classification and recognition due to its excessive accuracy. The CNN follows a hierarchical model which struggles on
constructing a network, form of a funnel, and eventually offers out a definitely related layer the place all the neurons
are linked to each and every different and consequently the output is processed A computer learns to classify images,
text and sound. The pc is trained with large image datasets then it changes the pixel value of the image to an indoor
representation, where the classifier can detect patterns on the input image.
We proposed a model that uses CNN network to classify Images between Human and Dogs. In the Proposed model
we are using a model VGG16 Net to find the predicted output for a given image.
II. LITERATURE SURVEY
This paper presents a solution for identifying dog breeds using their images of their faces. The proposed method
applies a deep learning based approach in order to recognize their breeds. [1]
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 753
[Link]
ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
This paper addresses the problem of biometric identification of animals, specifically dogs. They apply advanced
machine learning models such as deep neural networks on the photographs of pets in order to determine the pet
identity. In this paper, we explore the possibility of using different types of “soft” biometrics, such as breed, height, or
gender, in fusion with “hard” biometrics such as photographs of the pet’s face [2]
In this exertion, we define a CNN based approach for spotting dogs in perchance complex images and due to this fact
reflect inconsideration on the identification of the one of kinds of dog breed. The experimental outcome analysis
supported the standard metrics and thus the graphical representation confirms that the algorithm (CNN) gives good
analysis accuracy for all the tested datasets.. [3]
In this paper, we implement an android application that identifies the breed of a dog via image analysis, using a
Convolutional Neural Network (CNN) and transfer learning model. The android application lets the user click or
upload a picture of a dog. It then pre-processes the image and extracts the features required for testing. Prediction of
dog breed is done using CNN and transfer learning. [4]
In this paper ADA boosting methodology is used for breed analysis and recognition. ADA Boosting creates a strong
classifier from several weak classifiers. To separate the dog breeds from one another, we use Image processing
classification. It predicts the predominant breed/s present in the canine with maximum accuracy. [5]
This project uses computer vision and machine learning techniques to predict dog breeds from images. First, we
identify dog facial key points for each image using a convolutional neural network. These key points are then used to
extract features via SIFT descriptors and color histograms. We then compare a variety of classification algorithms,
which use these features to predict the breed of the dog shown in the image [6]
The paper presents the classification methods for dog breed classification using image processing approaches
conventional based approaches by Local Binary Pattern (LBP) and Histogram of Oriented Gradient (HOG). [7]
In this project we are using various pre-trained models like VGG16, Xception, InceptionV3 to train over 1400
images covering 120 breeds out of which 16 breeds of dogs were used as classes for training and obtain bottleneck
features from these pre-trained models. [8]
The current paper presents a fine-grained image recognition problem, one of multi-class classification, namely
determining the breed of a dog in a given image. The presented system employs innovative methods in deep learning,
including convolutional neural networks. Two different networks are trained and evaluated on the Stanford Dogs
dataset. The usage/evaluation of convolutional neural networks is presented through a software system. [9]
This paper considers application of machine learning in the context of animal identity management of veterinary
practice. In this application, electronic medical records of animal s would include digital photo graphs that are used to
identify them using image processing and recognition technologies. We investigated how combination of the " soft"
biometrics such as breed, a s well as face biometrics of dogs can improve identification of dogs. We apl)ly transfer
learning on GoogLeNet to perform the breed classification on the proposed BreedNet, and then to identify individual
dogs within the classified breeds, on the proposed DogNet.[10]
These were overall paper ideologies we went through and they would help us to make our system to work in a great
manner.
III. PROPOSED SYSTEM
In this paper we propose a system thst helps identify and classify the various dog breeds across the world. Our
system uses a pretrained VGG-16 model and further trained Convoluntional Neural Networks to be more precise.
In the initial preprocessing, the raw data is collected and is prepared for the primary processing analysis line and
extracting required data from a large set of data. Later the filtered data is passed to feature extraction, in this the
features of data like eye, face structure, mouth, hair are extracted. Further we have created 3 convolution layers with
relu activation function for extraction of features from image. Lastly, we have trained and validated the model.
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 754
[Link]
ISSN (Online) 2581
2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
IV. DISCUSSION
The fundamental idea of our discussion is to build a model that can predict breed of the dogs irrespective of its age.
Allocating data for training, validation and testing.
Formatting image input as the requirement of model.
Creation of augmented data
Convolutional Neural Network
Transfer Learning
4.1 Data Partition
The
he dataset has been divided into three parts train, validation and test for their specific purposes. Around 55% of the
dataset has been utilized into the training, 10% in validating and remaining data is allotted to testing. Allocating more
data towards training helps model learn better. The dataset includes various dog breed images along with their
respective names.
4.2 Data Preprocessing
Data preprocessing describes any type of processing performed on raw data to prepare it for another processing
procedure. The images have been resized into 224 * 224 pixels as it is required pixel size for inputs of VGG
VGG-16 model.
Then resized images are converted into tensor
tensors as per model requirements.
4.3 Creation of Augmented Dataset
The need for an augmented dataset is created due to insufficient dataset size. For augmented dataset vertical and
horizontal transformation are used. So the training of the model can be done on various version of single image which
are obtained through augmentation. Thus model becomes
becomes strong enough to identify dog breed in the images rotated in
any direction.
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 755
[Link]
ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
4.4 Convolutional Neural Network
Convolutional Neural Network is an exceptional kind of bolster advancing artificial neural network, which
stimulated through visual cortex. In CNN, the neuron in a layer is solely related to a small area of the layer earlier than
it, alternatively of all the neurons in a entirely related manner, so CNN manage fewer quantities of weights and much
less number of neurons. A CNN architecture is created from scratch with 3 convolutional layers and max pooling along
with dropout to dimensiontionallity and feature reduction.
4.5 Transfer Learning
In this experiment we used ResNet50 a pretrained neural network model as CNN architecture.ResNet50 model is
been trained on a large dataset called ImageNet. In this we freeze the early layers which contains the general
characteristics and added a fully connected output layer which gives us 133 classes(dog breeds) as an output.
V. RESULTS
This section discusses the obtained outcomes of our system and compares the model performance with the other
options explored.
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 756
[Link]
ISSN (Online) 2581
2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
5.1 Model Training
We then trained using these images on our constructed Neural Network with the architecture. Example Output of a
training epoch is shown below:
First we trained the CNN model from scratch and got accuracy of 12%. So to increase it, we opt to use transfer
learning. As ResNet-50 iss trained on larger dataset on ImageNet,
I we use it in transfer learning and gained an accuracy
of 70%.
[Link]. Algorithm Accuracy
1. CNN from scratch 12%
2. Transfer Learning using ResNet50 70%
5.2 Output
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 757
[Link]
ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
VI. CONCLUSION
In this system Convolutional neural network is a learning method for data analysis and predictions, now days it also
become very popular for image classification problems. Dog breed prediction of deep learning developed using
convolutional neural network is to predict the breed of hundred images in taking their images as input. Usage transfer
learning on the way to build model that make output and around to hundreds of dissimilar dog types. The results were
pretty good for the images the model was shown. The algorithm was able to identify dog breeds quite exactly. Transfer
learning takes an excessive choice in the upcoming in joining a pre-built model by the model we created.
REFERENCES
[1] Knowing Your Dog Breed: Identifying a Dog Breed with Deep Learning International Journal of Automation
and Computing 18(1), February 2021, 45-54
[2] Dog Identification using Soft Biometrics and Neural Networks Natural Sciences and Engineering Research
Council of Canada 22 July 2020
[3] Dog Breed Prediction using Convolutional Neural Network International Journal of Engineering and
Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-5, June 2020
[4] Breakthrough Conventional Based Approach for Dog Breed Classification Using CNN with Transfer 2019
11th Learning Information Conference on Information Technology and Electrical Engineering
(ICITEE),Pattaya, Thailand
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 758
[Link]
ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)
Volume 6, Issue 1, June 2021
Impact Factor: 4.819
[5] Dog Breed Classifier Using Deep Learning Jaypee University of Information Technology Waknaghat, Solan-
173234, Himachal Pradesh July 2019
[6] Dog Breed Identification with Fine tuning of Pre-trained models International Journal of Recent Technology
and Engineering (IJRTE) ISSN: 2277-3878, Volume-8, Issue-2S11, September 2019
[7] Recognition of Dog Breeds using Convolutional Neural Network and Visual Geometry Group International
Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-1, October
2019
[8] Dog Breed Identification Using Convolutional Neural Networks on Android CVR Journal of Science and
Technology, Volume 17, December 2019
[9] Using Convolutional Neural Networks to Classify Dog Breeds Hsu, David Stanford University 2018
[10] Transfer Learning on Convolutional Neural Networks for Dog Identification Xinyuan Tu , Kenneth Lai ,
Svetlana Yanushkevich 2018 IEEE
[11] Dog Breed Identification Using Convolution Neural Network and Web Scraping International Journal of
Science and Research (IJSR) ISSN: 2319-7064 ResearchGate Impact Factor (2018)
[12] Dog Breed Identification Using Deep Learning Zalan Raduly, Csaba Sulyok Babes, Zsolt Vadaszi
ResearchGate 2018
[13] Dog Breed Identification School of Computer Science University of Waterloo Waterloo 2017
Copyright to IJARSCT DOI: 10.48175/IJARSCT-1473 759
[Link]
