AI Certification | WorldQuant University

This project helps scientists track wildlife in Côte d'Ivoire by using deep learning models to classify animals in camera trap images. The workflow involves image preprocessing, fixing code issues, binary classification, and multiclass classification using CNN, RNN, and Backpropagation.

• 🎯 Objective: Classify images from camera traps to identify animals.
• 📂 Dataset: Images collected from Côte d'Ivoire wildlife reserves.
• 🛠 Tech Stack: Python, PyTorch, OpenCV, Scikit-learn.

The project follows a structured pipeline, with each step documented in separate notebooks:

📌 Notebook: 01_image_preprocessing.ipynb
✔️ Resizing & normalizing images.
✔️ Removing noise & unwanted artifacts.
✔️ Data augmentation (rotation, flipping, etc.).

📌 Notebook: 02_fixing_code.ipynb
✔️ Identifying & fixing coding errors.
✔️ Ensuring dataset integrity.
✔️ Verifying label consistency.

📌 Notebook: 03_binary_classification.ipynb
✔️ Model: Convolutional Neural Network (CNN).
✔️ Training: Binary classification - "Animal" vs. "No Animal".
✔️ Evaluation Metrics: Accuracy, Precision, Recall, F1-score.

📌 Notebook: 04_multiclass_classification.ipynb
✔️ Model: CNN + RNN (for sequential image patterns).
✔️ Backpropagation to optimize the network.
✔️ Fine-tuning: Transfer learning with ResNet/EfficientNet.
✔️ Handling imbalanced classes with data augmentation.