This repository showcases a collection of projects exploring various deep learning architectures and techniques across image classification, object detection, image generation, and natural language processing. Each project emphasizes different aspects of deep learning, from understanding model architectures to preprocessing data and evaluating model performance.
- Objective: Explore CNN architectures for classifying images into categories (airplanes, buses, cats, dogs, pizzas).
- Insights: Demonstrated the impact of padding and depth on classification accuracy.
- Objective: Develop a custom dataset for object detection.
- Insights: Quality data curation is crucial for object detection models' performance.
- Objective: Implement the YOLO model for object detection.
- Insights: YOLO efficiently processes real-time data, with challenges in detecting small objects and reducing false positives.
- Objective: Generate realistic images using GANs, focusing on pizza images.
- Insights: GANs can produce high-quality images, with architectural variations affecting image fidelity.
- Objective: Conduct sentiment analysis using GRUs.
- Insights: GRUs are effective in capturing temporal dependencies in text data for sentiment classification.
- Objective: Apply ViTs for image classification and compare their performance against CNNs.
- Insights: ViTs show promise in leveraging global context for classification, though they may require more data and resources.
Navigate into each project's directory to find detailed instructions on running the experiments, including data preparation, model training, and evaluation.
- Python 3.8+
- PyTorch 1.7.1+
- torchvision 0.8.2+
- Other dependencies listed in
requirements.txt
of each project.