International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 05 | May 2023
p-ISSN: 2395-0072
www.irjet.net
Dog Breed Identification Arshdeep Singh Ghotra1, Harashleen Kour2, Anas Hasan3 and Akash Khan4 1,2,3,4 Department of computer science and engineering, Apex Institute of Technology Chandigarh University,
Mohali,Punjab,India ---------------------------------------------------------------------***--------------------------------------------------------------------The application will utilize machine learning algorithms, Abstract - Dog Breed Identification has become essential to
such as TensorFlow, to analyze the photo and determine the breed of the dog in real time. TensorFlow plays a very important role in the project as it helps to write fast Deep learning code. It can run on a GPU (Graphics Processing Unit). GPUs are commonly used for deep learning model training and inference. As the dataset is very huge and there are many images it would take a lot of time to train, so we will use GPU which is 30 times faster than CPU in processing.
understand the conditions or climate in which dogs can survive. To identify dog breeds according to their physical features such as size, shape, and color, Dog Breed Identification techniques have been used. We have considered a dataset of 120 dog breeds to identify a dog's breed. This method begins with Convolutional Neural Networks (CNNs) or transfer learning. This method is evaluated with evaluation metrics and accuracy. And to achieve the best evaluation, we have also made use of Hyperparameter Tuning. In the deployment phase, we connected our model with the web Framework using Flask.
2. LITERATURE SURVEY Borwarnginn, et al. [4] proposed an approach to dog breed classification using transfer learning techniques. By leveraging pre-trained CNNs from large datasets such as ImageNet, the model was able to be trained with a small dataset. The proposed method uses deep learning and image augmentation to accurately identify dog breeds based on their face images. It was experimented with three different CNN models, namely MobilenetV2, InceptionV3, and NASNet. The results show that the NASNet model trained on a set of rotated images achieves the highest accuracy of 89.92%.
Key Words: Convolutional Neural Networks (CNNs), TensorFlow, GPU, MobilenetV2, Ngrok.
Flask,
Transfer
Learning,
1.INTRODUCTION Most of us have some liking for animals and the most liked animal of them is Dog [1]. Dogs are known for their loyalty, sweetness, and playfulness but on the contrary, some of them are dangerous too. We often encounter them in our daily routines, be it on the streets, in parks, or cafes. However, identifying the breed of a dog based on its appearance can be challenging, especially for those who are not well-versed in the different breeds. So, our project is based on identifying the breed of dog, which will help dog lovers to know which breed of dog will be suitable for the region where they live. It will be good for them as well as for dogs too. Because many dogs are not able to habituate themselves and may die at an early age. This project focuses on developing an app that provides a simple, fast, and reliable way to identify a dog’s breed through Image analysis and Convolutional Neural Network (CNN) architecture [2]. Analyzing Images using different computer techniques with predictive analysis that are being used in many different fields not only technology but agriculture too [3]. The application will be accessible through modern web browsers on desktop and mobile devices, making it easy for users to access it anywhere.
Uma, et al. [2] focused on fine-grained classification of dog breeds and the outcomes of the suggested system based on a large number of breeds. While the results demonstrate the potential of CNNs for predicting dog breeds, further research is required to investigate their efficacy. However, it is worth noting that the training times for neural networks can be quite lengthy, limiting the number of iterations possible within the scope of this study. Kumar, et al. [6] proposed an approach using OpenCV and the VGG16 model, which was successful in detecting human and dog faces and determining the corresponding breed using a combination of CNN and ResNet101 architecture. The model's performance exceeded expectations, achieving an accuracy of 81 percent compared to just 13 percent for a CNN model built from scratch. The results suggest that this approach holds significant promise for future research in the field of dog breed classification. Zhang, et al. [5] primarily focused on creating a cat detection model using deep learning techniques and deploying it through a mobile application. The application has been programmed to recognize 14 different types of cats, achieving an average accuracy rate of 81.74%. By optimizing the dataset and adjusting the hyperparameters, the model
Punyanuch Borwarnginn, et al. [4] trained the model which could be trained on a small dataset. They trained their model with 3 different CNN techniques, namely MobilenetV2, InceptionV3, and NASNet. Xiaolu Zhang, et al. [5] created a cat detection model using deep learning techniques and deploy it through a mobile application.
© 2023, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 1169