ELEC-475-Computer-Vision-with-Deep-Learning

Deep learning methods are highly effective at solving many problems in computer vision. This course serves as an introduction to these two areas and covers both the theoretical and practical aspects required to build effective deep learning-based computer vision applications.

Lab 1 MLP-4 Autoencoder:

This project focused on building and testing an MLP Autoencoder using the MNIST dataset. The project guides through the entire process from PyCharm setup to advanced autoencoder applications like image denoising and bottleneck interpolation. It features comprehensive instructions for implementing and training the autoencoder, as well as detailed steps for evaluating its performance on image reconstruction tasks.

Lab 2 Neural Style Transfer with AdaIN:

This lab from the ELEC 475 course, offers a comprehensive study of the AdaIN technique for neural style transfer. It includes a detailed exploration of network architecture, implementation, and performance analysis, with practical applications using diverse image datasets. The project demonstrates the nuances of style transfer, hyperparameter optimization, and the advantages of local GPU-based training.

Lab 3 Image Classification:

This next lab from the ELEC 475 course, showcasing the development and analysis of two CNN models for classifying images from the CIFAR-100 dataset. It includes a detailed exploration of VGG-like and ResNet-like architectures, training procedures, hyperparameter optimization, and performance evaluation, emphasizing the impact of architectural enhancements on classification accuracy

Lab 4 YODA; You Only Need Anchors:

Lab 4 introduces the concept of focusing on car detection using a ResNet-18 model. The project demonstrates the model's high accuracy in classification tasks and explores the complexities of object localization within images, as evidenced by a variety of qualitative examples and IoU analysis.

Lab 5 Pet Nose Localization

This final project focuses on a model that accurately localizes pet noses in images. It includes the process of model selection, training, and testing, along with detailed accuracy analysis and performance discussions, showcasing the practical application of computer vision techniques.