Use of CNNs to identify Wildlife Animals

We make use of Convolutional Neural Networks to identify wildlife animals.

Basic Steps

Step 1: Import Libraries and Load the Dataset (Imagefolder,...,Loaders)
Step 2: Create a CNN to Classify Wild Animals (from Scratch)
Step 3: Create a CNN to Classify Wild Animals (using Transfer Learning)
Step 4: Find Results

Dataset

In notebook we will be working with the Oregon Wildlife dataset created by David Molina with a google scrapper.It constains about 14.000 pictures of 19 different wildlife species. Count of picture by animal.

• 660 elk images.
• 696 bobcat images.
• 686 cougar images.
• 748 bald_eagle images.
• 717 canada_lynx images.
• 668 gray_fox images.
• 736 coyote images.
• 735 columbian_black-tailed_deer images.
• 718 black_bear images.
• 764 deer images.
• 577 mountain_beaver images.
• 728 virginia_opossum images.
• 726 sea_lions images.
• 701 nutria images.
• 759 red_fox images.
• 728 raccoon images.
• 656 raven images.
• 698 seals images.
• 730 gray_wolf images.
• 588 ringtail images.

Results

We achieve a high accuracy of 83% using our CNN and 92% with transfer learning.

This was a very fun project. Check the result in a image.