Master the full scope of KOAI Subject 3 (Computer Vision) hands-on. You'll learn to pick and use pretrained models from CNNs to diffusion models, and reach a level where you can train and evaluate classification, detection, segmentation, and generation tasks yourself. This is a high-school-division advanced course for students who have completed F1·F2.
Published: May 16, 2026 | Last updated: May 16, 2026 · Based on the KOAI 2026 guidelines
Track
Advanced
High-school division competitors
Target
High-school division
F1·F2 completed
Recommended Hours
About 10 hours
For 1:1 · varies 6–14 hours
KOAI Mapping
Full scope of Subject 3
Syllabus 3-1
By the end of A1, the student will master the full scope of KOAI Subject 3 (Computer Vision) hands-on. The goal is to reach a level where you can pick and use pretrained models from CNNs to diffusion models and train and evaluate classification, detection, segmentation, and generation tasks yourself.
After understanding how convolutional layers work in theory, you'll do transfer learning with pretrained encoders like ResNet and EfficientNet, detect objects with YOLO and DETR, and segment images with U-Net. Going all the way through CLIP, self-supervised learning, GANs, and Stable Diffusion, you'll conquer the full range of vision with hands-on code. An advanced trackIt is.
Below is the standard plan for 1:1. Depending on the student's prior knowledge and pace, some weeks may be accelerated, compressed, or covered in more depth. Core tools: PyTorch, torchvision, timm, albumentations, Ultralytics YOLO, Hugging Face, diffusers.
| Week | Topic | Key Deliverable |
|---|---|---|
| 1 | Convolutional layers (theory + hands-on), image classification baseline | CIFAR-10 CNN |
| 2 | Pretrained vision encoders (ResNet, EfficientNet) + transfer learning | Transfer learning notebook |
| 3 | Image augmentation (torchvision, albumentations) | augmentation ablation |
| 4 | Object detection: YOLO, SSD, DETR | YOLO fine-tuning |
| 5 | Image segmentation: U-Net | Medical and satellite image segmentation |
| 6 | Vision-text encoder: CLIP | zero-shot classification |
| 7 | Vision self-supervised learning (SimCLR/DINO concepts) | self-supervised embeddings |
| 8 | GAN image generation | DCGAN training |
| 9 | Diffusion models (using Stable Diffusion + theory) | Text→image generation |
| 10 | Capstone: real-dataset CV project | repo + demo |
※ Weeks are content units; actual time varies by student. Recommended about 10 hours, ranging 6–14 hours.
You'll write a Jupyter notebook each week. You'll train and evaluate classification, detection, segmentation, and generation tasks yourself, and by annotating the English notebooks with key Korean terms, you'll also prepare for the KOAI Round 2 Korean written exam at the same time.
You'll complete one ISEF/competition-grade CV project with a real dataset. Packaged as a repo and demo, this result becomes a core pillar of your KOAI portfolio.
A1 recommends 1 GitHub repo (koai-cv-projects) and 1 Huggingface Spaces demo. The capstone CV project becomes a core pillar of your KOAI portfolio, building up—as an accumulated asset—the kind of depth you can't manufacture in the final weeks before the exam.
GitHub
1 organized repo koai-cv-projects
Huggingface
1 Spaces demo (live demonstration)
Capstone
One ISEF/competition-grade CV project
The capstone CV project is the depth behind the 40% Round 1 document-review portfolio—the core deliverable for KOAI. The earlier you start, the more depth you'll have by the time you compete.
A1 is the first course in the advanced track of the KOAI curriculum. For the full track structure, see KOAI Prep Curriculum Hub.
Current Course
A1. Advanced I
Computer Vision
You need to have completed F1·F2 (or have equivalent skills). Recommended for students planning to compete in the high-school division.
Yes. A1 (CV) and A2 (NLP&Audio) can run in parallel, and both are advanced work in KOAI Subjects 3 and 4.
It's the full scope of Syllabus Subject 3 (Computer Vision) (3-1). It covers CNNs, transfer learning, object detection, segmentation, CLIP, GANs, and diffusion.
It's one ISEF/competition-grade CV project, and it becomes a core pillar of the KOAI Round 1 document-review portfolio. We recommend a Huggingface Spaces demo.
The order is A2 (NLP&Audio) → C1 Portfolio Studio → C2 Mock Bootcamp → C3 Selection Camp. For exact dates, check the KOAI competition guide (https://citcoding.com/competitions/koai.html).
We assess your F1·F2 completion level and design a personalized A1·A2 advanced path for you.