Streamlit: Turn Data into Apps

Advanced Image Classification and Object Localization System

Students will design, implement, and evaluate a Convolutional Neural Network (CNN) architecture from scratch or using transfer learning (e.g., ResNet, EfficientNet) to solve a complex visual recognition task. The project involves comprehensive data preprocessing, data augmentation strategies, hyperparameter tuning, and a detailed performance analysis using metrics like precision-recall curves and visual interpretability tools such as Grad-CAM. The final submission must include a technical report and a functional Jupyter Notebook containing the full training and inference pipeline.