ANIMAL CLASSIFICATION USING DEEP LEARNING
Shweta Suryawanshi, Vaishali Jogdande, Ankita Mane · International Journal of Engineering Applied Sciences and Technology · 2020
Kaggle's Dogs vs. Cats contest is trying to solve the CAPTCHA challenge, which is based on the issue of distinguishing dog and cat images.It's easy for humans, but evidence suggests that the automatic separation of cats and dogs is particularly difficult.Many people have been working or working on building classifiers for machine learning to address this issue.A color-based classifier had an accuracy of 56.9% on the Asirra dataset.A SVM classifier achieved an accuracy of 82.7 percent based on a combination of color and texture characteristics.And in, they used the features of SIFT (Scale-Invariant Feature Transform) to train a classifier then finally got a 92.9 percent accuracy.We also want to solve this problem in our plan and achieve higher efficiency.We've tried various strategies.We tried Dense-SIFT features, combining Dense SIFT and color features, and features learned from CNN, for example.We also used SVMs on the learned features and finally achieved 94.00 percent of our best classification accuracy.