█ class-only course site
CMSC 178Image Processing (IP)
Teach a computer to see. Ten days from "an image is a grid of numbers" to a convolutional neural network you trained, tested and can defend.
Made by Eusib Vincent Pulvera
for the CMSC 178 class.
Log in with your class email → See the 10 days
No password. We email you a one-time code. Only students added by the instructor can enter.
How each day works
Read
Short pages, one idea at a time. Every line of code is explained beside it.
Type
Type the key code yourself. No pasting. That is how it sticks.
Run
Run the notebook on your own computer and watch real results appear.
Prove it
A practice quiz and a short write-up show you actually understood.
The 10-day roadmap
-
DAY 1
Setup and images as tensors
An image is a grid of numbers. -
DAY 2
ML fundamentals and your first network
Train a tiny model by hand. -
DAY 3
Convolution, built by hand
Slide a kernel; find the edges. -
DAY 4
Your first CNN in Keras
Build it, train it, look inside. -
DAY 5
Data pipelines
Augmentation, leakage, imbalance. -
DAY 6
Training well, evaluating honestly
Seeds, intervals, one look at the test set. -
DAY 7
Transfer learning
Borrow a pretrained network. -
DAY 8
Reproducibility and your group's experiment
Run a mini version of your study. -
DAY 9
Writing up the research
Claims that match the evidence. -
DAY 10
Final assessment and certificate
Safe Exam Browser. One attempt.
Your group's research
Six groups, four tracks: domain shift, subgroup audit, compression and degradation. By Day 8 you run a mini version of your own study.
The final, and a certificate
Day 10 is taken in Safe Exam Browser: one attempt, 80% to pass. Pass and you get a certificate that anyone can verify online.
CMSC 178 Image Processing (IP)
Made by Eusib Vincent Pulvera for the class