Dog breed classification using convolutional neural network and transfer learning

Manasi Koli , Mansi Chalawari , Mrunal Deshmukh , Poonam Pathak

Convolutional Neural Network, Transfer learning, ResNet50, Dog Breed Classification, Deep Convolutional Neural Networks.

Convolution Neural Networks (CNN) with the help of transfer learning the system used for dog breed classification and detection. It also considers the combination of multiple basic CNN models for dog breed can be classified and give immense improved performances. This model supports the classification of 133 dog breeds whereas capital punishment the model. CNN may be a structure designed to method real-world user-defined pictures. Consider, a dog image is chosen the model can classify Associate in Nursing establish an analysis of the dog breeds, this model is used as a neighborhood of mobile or net app for the real-world and user-definition image. once the image is given to the model for execution, it analyzes if a dog is a gift and returns the estimated breed. during this model, we tend to use a pair of datasets that are dog and human. whereas we tend to execute our scratch model it’s impossible to induce high accuracy therefore we tend to come up with the Resnet50 model that has high accuracy share up to eighty-six and higher than roughly that processes information at intervals less time with the assistance of GPU (Graphical Process Unit). The result that occurred has shown Brobdingnagian improvement of accuracy as compared to the scratch model.

Volume 6 Issue 4, April-2021

Pages : 99 - 103