Assignments

Intelligent Vision Systems · BSES, 7th Semester, Section A · University of Lahore · Fall 2026

Four graded assignments plus periodic quizzes make up the 10-mark Quizzes/Assignments component of the course grade (see the Syllabus). Assignments are individual unless stated otherwise; out/due dates also appear on the Schedule page.

AssignmentDescriptionOut / Due
Assignment 1
Image processing		 Apply point operations, histogram equalisation/CLAHE and thresholding to a supplied dataset; compare methods quantitatively. Out Week 3 · due Week 4
Assignment 2
Finding lane lines		 Build a self-driving-car lane-finding pipeline — colour selection, Gaussian blur, Canny edges, region mask and Hough lines — then average and extrapolate to one line per lane on six test images and two videos, and write up its shortcomings (Lecture 5 Part II, Lab 5). Out Week 6 · due Week 7
Assignment 3
Motion & tracking		 Implement optical flow and a tracking-by-detection pipeline; evaluate with simple MOT metrics on a short clip. Out Week 11 · due Week 12
Assignment 4
Deep detection		 Train a custom YOLO detector on a chosen dataset, report mAP, and write a deployment note for the Raspberry Pi 5. Out Week 13 · due Week 14

Quizzes: short in-class quizzes at the start of selected sessions test the previous week's theory (best scores counted).

Assignment hand-outs, starter files, and rubrics will be posted here as each one goes out.

Assignment 2 — Finding Lane Lines on the Road

Out: Week 6 (Mon 5 Oct 2026) · Due: before the Week 7 session (Mon 12 Oct 2026) · individual · companion to Lecture 5 Part II — Computer Vision for Self-Driving Cars (PDF) and Lab 5.

Write a pipeline that identifies the lane lines on the road, first in six test images and then in two video streams. Adapted from Project 1 of the Udacity Self-Driving Car Engineer Nanodegree (udacity/CarND-LaneLines-P1, MIT licence).

Open the starter notebook in Colab View on GitHub Download notebook Write-up template

Tasks & marks

PartTaskMarks
1Pipeline. Implement process_image() from the lecture tools (grayscale, Gaussian blur, Canny, region of interest, Hough lines, overlay) and run it on all six test images; show intermediate results for one image.30
2One line per lane. Rewrite draw_lines() to separate segments by slope, average them and extrapolate a single solid line for the left and right lane from the bottom of the image towards the horizon.30
3Video. Process solidWhiteRight.mp4 and solidYellowLeft.mp4; the lines must stay on the lanes for the whole clip with little jitter (e.g. smooth slope/intercept across frames, reuse the last good line when none is found).20
4Write-up. Describe the pipeline and your parameter choices with evidence, list honest shortcomings, and suggest improvements.20
★Bonus. Make the pipeline work on challenge.mp4 (curve, shadows, bright concrete, 1280×720).+10

What to submit

  1. 05-lane-finding-project.ipynb with all cells run and outputs visible.
  2. Your output videos: test_videos_output/solidWhiteRight.mp4 and test_videos_output/solidYellowLeft.mp4 (and challenge.mp4 if attempted).
  3. The write-up (last notebook cell or writeup.md based on the template).

Marking guide

  • Excellent (85–100%) — one stable, well-placed line per lane on every image and throughout both videos; resolution-independent region; parameter choices justified with intermediate images; insightful, specific shortcomings and improvements.
  • Good (65–84%) — lanes found on all images and most video frames, with occasional jitter or a short drop-out; reasonable write-up.
  • Satisfactory (40–64%) — working pipeline, but lines are segmented or unstable on video, or the write-up is generic.
  • Insufficient (<40%) — pipeline incomplete or does not run.

Academic integrity: many solutions to this Udacity project are online. Write your own code and be ready to explain any line of it in a short viva; code you cannot explain earns no marks. Cite any external source you used.