Daalakker 9, 2200 Herentals
Cat Recognition Deep Learning Project
Project Overview
This deep learning project, built using Keras and TensorFlow, aimed to classify five different types of cats—domestic cat, lion, tiger, cheetah, and leopard—through image recognition. The project encompassed all stages of a typical deep learning workflow, from data collection to model training and prediction.
Key Components
- Data Collection: A dataset featuring images of the five cat species was compiled from various sources.
- Exploratory Data Analysis (EDA): EDA was performed to understand the dataset, visualize image distributions, and address any class imbalances.
- Data Preprocessing: Images were resized, and the dataset was split into training, validation, and test sets. Data augmentation was used to enrich the training data.
- Model Training: Ten different models, including Xception, ResNet, Inception, and MobileNet, were fine-tuned using transfer learning. The models were trained and validated on augmented data.
- Model Evaluation: Models were evaluated using accuracy, loss, and confusion matrices, allowing selection of the best-performing model.
- Prediction: The final model was tested on a separate dataset, and predictions were stored in a CSV file for further analysis.
Technologies Used
- Keras
- TensorFlow
- Transfer Learning
How to Run Locally
- Clone the repository to your local machine.
- Explore the Jupyter notebook to follow the workflow and insights.
- Train the models using the provided notebooks.
- Evaluate the models and make predictions on test data.
