Welcome to HappyPaws ! This project utilizes a sophisticated machine learning model to analyze and classify dog emotions into four categories: happy, angry, sad, and relaxed, with an impressive accuracy rate of about 78%.
The model has been trained on a wide range of images to ensure high accuracy and reliability. During data preparation, an advanced object recognition model was used to filter the images. Only images with a confidence threshold over 90% were selected to ensure they clearly featured dogs. Any images with people were excluded to keep the focus solely on the dogs.
After detecting the dogs in the images, the photos were cropped to center on the dog, minimizing the background and highlighting the dog's features. This careful data cleaning process significantly improved the model's accuracy.
As a result of these careful methods, the model can correctly identify a dog's emotional state based on their photo in nearly 8 out of 10 cases.