Computer Vision

What you'll learn

• Explain image formation, sensors, color and sampling, apply the pinhole camera model, and use convolution for filtering and edge detection
• Apply point operations, color spaces, morphology, edge, line and corner detection, feature matching, homographies and RANSAC to classical vision problems
• Design and analyze convolutional neural networks, computing output sizes, parameters and receptive fields, and train and adapt them with augmentation and transfer learning
• Select efficient architectures for edge devices, build leakage-free datasets, evaluate classifiers with appropriate metrics, and calibrate model confidence
• Build and evaluate object detection and segmentation systems using IoU, non-maximum suppression, average precision, mean IoU and Dice, and assess the use of face recognition
• Explain assignment, focal loss, IoU-based losses and DETR set prediction, and apply panoptic segmentation, pose estimation and tiling for small objects
• Apply vision transformers, self-supervised learning and contrastive vision-language models such as CLIP, and analyze their costs and limitations
• Explain VAEs, GANs and diffusion models, including latent diffusion and guidance, evaluate generators with FID, and assess synthetic data and deepfake risks
• Estimate motion, track objects and recover 3D structure with optical flow, stereo, structure from motion and depth sensors, and reason about safety-critical vision systems
• Calibrate cameras and correct distortion, apply epipolar geometry, triangulation, PnP and bundle adjustment, and describe NeRF and Gaussian splatting
• Build and evaluate vision-language systems, including open-vocabulary detection, multimodal language models, OCR and document understanding
• Design, size, monitor and secure deployed vision systems, protect privacy and fairness under applicable law, and deliver a complete capstone project