Abstract:
This is a deep learning approach to recognizing animals. We have used images of various
domesticated and wild animals. The deep learning procedure instructs machines to detect
and bring out the features from images. We believed that we would be able to build a
model which could analyze animals and give a more efficient outcome. Firstly we have
collected images to create datasets and completed preprocessing. Then we applied
various processes to get the result and then applied Deep CNN and trained the dataset.
Then completed the detection process of the animal (Example: Elephant). For this
approach, we have used python. Implementation is done on a previously trained model
EfficientNet B5. To give the best and true outcome with higher accuracy, Deep CNN
formatted the image dataset in several ways. The training accuracy of the image is around
94.29%. The CNN model classified most of the image dataset of animals where only 118
images were misclassified. We got a validation accuracy of around 93.95%. The testing
accuracy was around 93.45%.In this animal detection approach, we have used algorithms
for classifying animals. We have used a huge amount of images for this purpose. To get
more accurate results, we recommend using more data. Working with a lot of data has the
potential to be 100% accurate. Another thing is that this study can be done based on
animal voices.