This project implements a deep learning-based image classification model to classify animals into various categories using Convolutional Neural Networks (CNN), ResNet50V2, and MobileNet for feature extraction.
We used the Animals-10 Dataset, a publicly available dataset on Kaggle that contains images of 10 different animal species.
📌 Dataset Source: Animals-10 Dataset on Kaggle
The dataset consists of ~26,000 images categorized into 10 animal classes, originally labeled in Italian.
We renamed them to English for consistency.
- Custom CNN Model trained for image classification
- MobileNet Transfer Learning model for improved accuracy
- ResNet50V2 for high-performance classification
- Multi-class classification with 10 animal categories
- Handles a dataset of 26,000+ images
We trained three different models on the dataset and obtained the following accuracies:
| Model | Accuracy |
|---|---|
| CNN | 67% |
| MobileNet | 94% |
| ResNet50V2 | 90% |
- MobileNet performed the best due to transfer learning, achieving 94% accuracy.
- ResNet50V2 also showed strong performance with 90% accuracy.
- The CNN model, trained from scratch, achieved 67% accuracy.
Due to the large size of the ResNet50V2 model and presentation file, they are hosted on Google Drive. You can access them using the links below:
- ResNet50V2 Model: Download Here
- Presentation: Download Here
📦 Deep Learning Animal Classifier
│-- 📜 README.md - Project documentation
│-- 📜 Homepage.py - Main script for running the app
│-- 📜 Deep Learning Animal classifier.ipynb - Jupyter Notebook with training code
│-- 📜 CNN.h5 - Trained CNN model
│-- 📜 MobileNet.h5 - Trained MobileNet model
│-- 📂 test_image/ - Sample images for testing