Mohammad Sabokrou
Associate Professor
New Uzbekistan University
Computer Vision
From images as numerical arrays to learned representations, detection, segmentation, tracking, and responsible evaluation. Each topic pairs a two-hour lecture with an experiment-led lab.
16 weeks2h lecture + 2h labCS & AIGoogle Colab
m.sabokrou@newuu.uz
Email before an office visit to reserve a short time slot.
Measure, break, repair
Every lab uses evidence from code, including at least one failure case and an explicit evaluation step.
Student materials
Lecture
Lab manual
Lab 0
Lab 1
Open in Google Colab
Start here
Session 1: Images as Numbers
A visual 54-slide, two-hour lecture with real applications, AI-generated imagery, human and machine vision, image formation, image types, matrices, worked examples, and Lab 1 / Week 2 previews.
Open watermarked PDFStudent edition
Lab 0 plus fifteen two-hour labs, with objectives, timing, deliverables, expected results, and common pitfalls.
Open watermarked PDFGetting Set Up
A guided Colab notebook with beginner warm-ups, a four-problem Challenge Track for early finishers, open code cells, and a 16-point readiness self-test.
Download notebookImages as Numbers in Practice
Run the complete examples, inspect each image and result, make one small change, then explain your own evidence. No code writing is required.
Open student guideOpen in Google Colab
Submission rule. Email the executed notebook and short PDF report to m.sabokrou@newuu.uz by 23:59 on the same day as your scheduled lab. Use StudentID_WeekXX.ipynb and StudentID_WeekXX_Report.pdf; run all cells and keep outputs visible. Instructor solutions, marking rubrics, and internal production notes are intentionally not published. For Lab 11, review the current Ultralytics licensing terms before publishing or deploying a project.
16-week learning plan
Select any week to view its lecture focus, practical work, and available files.
01
Images as numbers
Arrays, channels, bit depth, nuisance factors, and the limits of pixel matching.
02
Light, colour, and histograms
Image formation, colour spaces, contrast, and histogram equalisation.
03
Filtering and edges
Convolution, kernels, separability, gradients, and edge detection.
04
Features and matching
Corners, local descriptors, robust matching, RANSAC, and panoramas.
05
Learning a classifier
Linear models, softmax, loss functions, gradients, and optimisation.
06
From filters to CNNs
Learned filters, convolutional layers, feature maps, and receptive fields.
07
Generalisation
Overfitting, augmentation, uncertainty across runs, and controlled comparisons.
08
Midterm and consolidation
Review the foundations, hand-designed features, learning, and generalisation.
09
Transfer learning
Pretrained representations, frozen features, fair baselines, and negative transfer.
10
Architectures and depth
Residual connections, vanishing gradients, latency, and Pareto trade-offs.
11
The detection problem
Bounding boxes, IoU, non-maximum suppression, precision-recall, and AP.
12
Detectors in practice
Annotation, training, validation, failure analysis, and label quality.
13
Segmentation
Dense prediction, U-Net structure, overlap metrics, and boundary quality.
14
Motion and tracking
Optical flow, the aperture problem, association, and identity switches.
15
Evaluation and error analysis
Confidence intervals, group metrics, model cards, compression, and audit design.
16
Transformers, ethics, and demos
Vision transformers, attention maps, failure cases, and responsible deployment.