CSE754: Deep Learning for Computer Vision

Fundamentals of Computer Vision: image representation, convolution, feature extraction, classical vs. deep learning approaches; Convolutional Neural Networks (CNNs): architecture, layers, pooling, backpropagation; Popular CNN Architectures: LeNet, AlexNet, VGG, ResNet, EfficientNet; Object Detection and Segmentation: R-CNN, YOLO, Mask R-CNN, semantic vs. instance segmentation; Vision Transformers (ViTs): self-attention in vision, Swin Transformer, hybrid models; Multimodal Vision-Language Models: CLIP, BLIP, Flamingo, LLaVA, aligning vision and text representations, cross-modal learning; Image Generation and Synthesis: GANs (StyleGAN, BigGAN), VAEs, Diffusion Models (Stable Diffusion, Imagen); 3D Vision: depth estimation, point cloud processing, 3D CNNs, 3D object recognition and segmentation, 3D shape representations, 3D shape reconstruction, NeRF; Video Analysis: transformers for video, action recognition; Self-Supervised and Few-Shot Learning: contrastive learning (SimCLR, MoCo), meta-learning; CUDA Deep Neural Network (cuDNN), Applications: medical imaging, autonomous driving, facial recognition, industrial product quality inspections, creative AI; Challenges and Ethics: adversarial attacks, fairness, bias, privacy concerns. Project: real-world problem solving based on computer vision techniques and deep learning algorithms.

Course Objectives

The core objectives of this course are to:
Understand the fundamentals of computer vision and deep learning
Acquire adequate knowledge - methods, algorithms, tools and techniques
Develop skills for 2D and 3D computer vision with the use of state of the art deep learning algorithms
Apply theoretical knowledge and technical skills of computer vision and deep learning to solve real-life problems

List of Books

1. To Be Added

Course Outcome