Elephants are detected from a given video with the help of frame extraction and convolutional neural networks
Image classification is performed on the frames extracted from a given video. The videos were collected from youtube which mainly consists of attacks in rural areas in the state of Kerela, India. Image classifiction is performed with the help of Convolutional Neural Networks. If at least 48 frames or images are classified as elephant in from the given video then the result declared as an elephant attack occuring. This is check if the elephant is visible for atleast 2 seconds and the videos are captured at 24 frames per second. This can reduce false postives and ensure improved accuracy of the model.