Course Schedule — Intelligent Vision Systems
BSES · 7th Semester · Section A · University of Lahore · Fall 2026
Each 3-hour Monday session pairs a 1.5-hour theory lecture with a 1.5-hour supervised laboratory on the Raspberry Pi 5. The schedule below is subject to change as the semester progresses; lecture topics may shift slightly with class pace. Lab, quiz, assignment and exam dates remain fixed except in extraordinary circumstances.
Semester: 31 August – 14 December 2026 (16 weekly sessions) · Meeting: Mondays, 11:00 AM–2:00 PM · E1-404 Lab. Dates assume a Monday start on 31 Aug 2026 — replace them with the official University of Lahore academic-calendar dates and insert public-holiday / mid-term-break rows as needed. Reading keys: S = Szeliski, Computer Vision 2e, G&W = Gonzalez & Woods, Digital Image Processing 4e.
| Wk | Date | Lecture & Theory Topics (1.5 h) | Laboratory — Raspberry Pi 5 (1.5 h) | Slides | Assessment / Reading |
|---|---|---|---|---|---|
| 1 | Mon 31 Aug | L1–L2 — Introduction & Image Formation I. What an intelligent vision system is; acquire → process → analyse → decide → act; edge/embedded constraints; the human visual system; optics & projection: radiometry, lenses, aperture, depth of field, pinhole & thin-lens models, perspective projection, exposure & motion blur. | Lab 1 — Raspberry Pi 5 & Embedded Systems I: OS install, headless SSH/VNC, Python env, GPIO & embedded-vision overview. | Lec 01 | Lab 1 out · S 1–2.2 |
| 2 | Mon 7 Sep | L3–L4 — Image Formation II & Digital Fundamentals. CCD vs CMOS; Bayer CFA & demosaicing; the ISP pipeline; rolling shutter; gain/ISO/read noise; HDR capture. Sampling & quantisation; resolution & bit depth; pixel neighbourhoods/connectivity; distance metrics; colour spaces (RGB, HSV, YCbCr, Lab). | Lab 2 — RPi 5 & Embedded II: install & verify OpenCV 5.0 (note key v4→v5 changes); Pi camera stack (libcamera/picamera2), image & video I/O on the Pi, capture FPS. | Lec 02 | Lab 2 out · S 2.3 · G&W 2.4–2.6 |
| 3 | Mon 14 Sep | L5–L6 — Deep Learning for Computer Vision. The perceptron; activation functions; building dense networks; loss functions; gradient descent & backpropagation; adaptive optimisers; mini-batch SGD; overfitting & regularisation (dropout, early stopping). TensorFlow & PyTorch intro labs. | Lab 3 — CV Fundamentals I: load/save images & videos, drawing functions, basic image operations. | Lec 03 | Lab 3 · Assign. 1 out · G&W 3.1–3.3 |
| 4 | Mon 21 Sep | L7–L8 — Computer Vision: How Computers See the World (with practical implementation). The semantic gap & why vision is hard; images as numbers (pixels, grayscale, RGB tensors). The ladder of tasks: image classification (edges → shapes → objects, learned detectors); object detection (sliding windows vs one-pass detectors, confidence threshold); segmentation (masks, semantic vs instance); image understanding (relationships, captioning, VQA with vision-language models). | Lab 4 — Computer Vision in Practice: image as numbers, ResNet-50 classification, RF-DETR detection & segmentation (background blur/replace), VQA with SmolVLM2 — five notebooks, one per lecture demo. | Lec 04 | Lab 4 · Assign. 1 due · S 1, 6.1–6.6 |
| 5 | Mon 28 Sep | L9–L10 — Computer Vision: Images, Colour Spaces & Filters. Human vs computer vision; pixels, grayscale 0–255, RGB channels & images as tensors. Colour spaces: RGB/BGR, HSV, LAB (lane finding), YCbCr, CMYK. Filters: convolution, averaging, Gaussian & median blur worked by hand; Sobel, Laplacian & Canny edges; shapes, contours & morphology. Feature extraction: keypoints, descriptors & ORB matching. Deep-learning-based CV: learned features, CNNs and Vision Transformers. Part II — Computer Vision for Self-Driving Cars: finding lane lines with colour selection (RGB, HLS), region-of-interest masks, Canny edges, the Hough transform, averaging/extrapolation and video. | Lab 5 — Self-Driving Car Vision: colour selection, region masking, Canny edges and Hough lines — four notebooks, then the lane-finding project (Assignment 2). | Lec 05 I Lec 05 II |
Lab 5 out · G&W 6.1–6.3, 3.5, 10.2 · S 2.3, 7.2, 7.4 |
| 6 | Mon 5 Oct | L11–L12 — Camera Calibration. Why distortion matters for measurement; the pinhole model, intrinsics K (focal length, principal point) and extrinsics [R | t]; radial (barrel, pincushion, mustache) and tangential distortion and the OpenCV coefficients; camera types — standard, wide-angle, fisheye, omnidirectional, stereo, RGB-D, rolling vs global shutter, Raspberry Pi modules; calibration targets (chessboard, circle grids, ChArUco). Calibrating with OpenCV: corners & sub-pixel refinement, object/image points, Zhang’s method, reprojection error, undistortion & remap; ChArUco, fisheye and stereo calibration; ROS 2 camera_calibration; perspective transforms, bird’s-eye view & lane curvature. (Filtering, frequency, restoration & morphology: supplementary deck.) |
Lab 6 — Camera calibration: pinhole & distortion, chessboard corners, calibrate & undistort a GoPro, dash-cam bird’s-eye view & curvature, ChArUco & fisheye, bug hunt, stereo rectification, calibrate your own Pi camera — eight notebooks. | Lec 06 | Lab 6 · Assign. 2 out · S 2.1, 11.1 · H&Z 6–7 · Zhang 2000 |
| 7 | Mon 12 Oct | L13–L14 — Edges/Contours & Local Features. Canny in depth; Hough transform for lines & circles; contour tracing; polygon approximation; region descriptors — moments, Hu invariants; shape matching. Harris & Shi–Tomasi corners; scale space; DoG blob detection; FAST; repeatability & invariance. | Lab 7 — Features & matching: corners/keypoints, ORB/SIFT descriptors, feature matching, Hough lines/circles. | Lec 07 | Lab 7 · Assign. 2 due · S 7.1–7.3 |
| 8 | Mon 19 Oct | L15–L16 — Descriptors, Matching & RANSAC. SIFT, SURF, ORB, BRIEF; ratio test & cross-check; brute-force vs FLANN; Hamming distance; why binary descriptors suit embedded targets. 2D transform families; homography by the DLT; robust fitting (RANSAC/MSAC); alignment; panorama stitching & blending. | Lab 8 — Geometry: homography + RANSAC, panorama stitching, camera calibration & undistortion. | Lec 08 | MIDTERM (20) · Lab 8 · S 8.1–8.2 |
| 9 | Mon 26 Oct | L17–L18 — Segmentation I & II. Region growing & split/merge; k-means & mean-shift in colour space; SLIC superpixels; IoU/Dice evaluation. Watershed; graph cuts & GrabCut; active contours (snakes); grouping cues; from hand-designed masks to learned segmentation (bridge to U-Net). | Lab 9 — Build a real-world CV system: plan & architecture, measure object size (2-D), defect-recognition system, database integration. | Lec 09 | Lab 9 · Project proposal due · S 7.5 |
| 10 | Mon 2 Nov | L19–L20 — Calibration & Stereo. Intrinsic/extrinsic parameters; lens distortion models; Zhang’s planar calibration; reprojection error; undistortion & rectification. Epipolar geometry; fundamental & essential matrices; rectification; block-matching & SGBM disparity; depth from disparity; RGB-D sensors. | Lab 10 — Calibration, stereo & depth on the Pi. | Lec 10 | Lab 10 out · S 11.1–12.1 |
| 11 | Mon 9 Nov | L21–L22 — Optical Flow & Tracking. Brightness constancy & the aperture problem; Lucas–Kanade & Horn–Schunck; pyramidal KLT tracking; dense flow (Farnebäck). Background subtraction (MOG2, KNN); the Kalman filter; mean-shift/CAMShift; correlation trackers (KCF, CSRT); tracking-by-detection & SORT; MOT metrics. | Lab 11 — Motion & tracking: optical flow, OC-SORT & Strong-SORT tracking, object trajectory & speed estimation, line-crossing counting. | Lec 11 | Lab 11 · Assign. 3 out · S 9.3–9.4 |
| 12 | Mon 16 Nov | L23–L24 — Classical Recognition & Neural Networks. Sliding-window paradigm; HOG + linear SVM (pedestrian detection); Haar cascades (Viola–Jones); bag-of-visual-words & spatial pyramids. Perceptron → MLP; loss functions; back-propagation; SGD, momentum, Adam; over-fitting, regularisation, augmentation; train/val/test discipline. | Lab 12 — Deep learning: image classification (transfer learning) + YOLO object detection on image/video/camera, custom training on Colab, SAHI, pose. | Lec 12 | Lab 12 · Assign. 3 due · Goodfellow 6–8 |
| 13 | Mon 23 Nov | L25–L26 — CNNs, Segmentation Networks & Reading Research Papers. Convolution & pooling; receptive fields; batch norm; LeNet → ResNet → MobileNet → EfficientNet; transfer learning; two-stage (Faster R-CNN) vs one-stage (SSD, YOLO) detection; semantic segmentation (FCN, U-Net, DeepLab); Mask R-CNN; mAP. How to read a research paper and reproduce a model architecture from scratch. | Lab 13 — Paper Implementation I: U-Net from scratch (PyTorch) — read the paper, build the encoder–decoder with skip connections, train for segmentation (IoU/Dice). | Lec 13 | Lab 13 · Assign. 4 out · Ronneberger 2015 · S 5.3 |
| 14 | Mon 30 Nov | L27–L28 — Edge Vision & the OpenCV 5.0 Deep Dive. Accelerators (GPU, NPU, DSP, FPGA); inference runtimes (TensorRT, TFLite, ONNX Runtime, OpenVINO, Hailo); memory hierarchy, DMA, zero-copy; latency/throughput/power; PTQ vs QAT, INT8/FP16, pruning, distillation, operator fusion. OpenCV 5.0 deep dive: modernised architecture; the new DNN engine; ONNX & modern DL workflows; hardware acceleration; breaking changes & migration from OpenCV 4; performance optimisation. | Lab 14 — Modern deployment on RPi 5: OpenCV 5.0 pipeline migration + CascadeClassifier fix; YOLOv8n ONNX detection with the new DNN engine; export to TFLite/ONNX, run on ONNX Runtime / Hailo, INT8 quantisation, benchmark. | Lec 14 | Lab 14 · Assign. 4 due · OpenCV 5 · TinyML |
| 15 | Mon 7 Dec | L29–L30 — Vision Transformers, Advanced Topics & Review. Vision transformers & self-supervised learning; multimodal & open-vocabulary models; event cameras; a SLAM overview; ethics, privacy, dataset bias, safety & robustness. Course wrap-up and final-exam review. | Lab 15 — Paper Implementation II: the Transformer (“Attention Is All You Need”) from scratch (PyTorch) — attention, multi-head, positional encoding, encoder/decoder; bridge to ViT. | Lec 15 | Project report due · Vaswani 2017 · S 5.5, 14.x |
| 16 | Mon 14 Dec | Capstone Demos & Review. Live demonstration of the capstone vision system on Raspberry Pi 5 target hardware; short talk + Q&A; peer review; final-exam review. | Capstone project demonstrations (Lab Project, 30 marks). | Lec 16 | Capstone demos · Final Exam (40, finals week) |
To attach material to a session, edit the matching row in schedule.html (and the corresponding entry in lectures/index.html) and drop in a link to the slide deck, notes, reading, or recording. Model and tool versions move quickly — reconfirm YOLO26, OpenCV 5.0, Hailo and library versions before each offering.