Objectives of this course:

Learn the basic methods of digital image analysis and pattern recognition: pre-processing, image segmentation, shape representation and classification. These concepts will be illustrated by applications in computer vision and medical image analysis.

2025 Schedule: Fridays, morning (8:15am - 12:00am)

Labs:

The lab materials will be available through the public GitHub repository: https://github.com/LTS5/iapr2025. Lab assignments are designed to develop practical implementation skills related to the topics covered in class and serve as preparation for the final project. This project will be a hands-on application of the concepts learned throughout the course. 

Labs are provided as Jupyter Notebooks and must be completed in groups of three students. Please form your groups on Moodle before March 7th. Note that the same groups will be used for the final project later in the semester.

We highly recommend consulting the

file in the GitHub repository for detailed instructions. This document includes essential information such as lab guidelines, Python and Jupyter installation steps, and best practices for using GitHub. Additionally, a tutorial is provided to help you get familiar with fundamental image processing functions in Python. The first lab will be released on March 7th. Please follow the provided instructions carefully to ensure a smooth and effective learning experience.

The Lab 1 is available on GitHub IAPR: https://github.com/LTS5/iapr2025
Use the same notebook to write your code and discussions. The routine(s) from the second part may be useful for the final project, so pay close attention.

Submission Details

• Work in groups of three and submit one lab per group.
• Deadline: March 19th at 23:59 (submission box below).
• The report will be graded—ensure all group members' names and SCIPER are included.
• Verify your code by rerunning the notebook before submission.
• Follow the naming convention: 
  lab_1_group_<group_id>.ipynb
• The notebook will be rerun with the provided libraries. No additional packages are allowed.

The Lab 2 is available on GitHub IAPR: https://github.com/LTS5/iapr2025
Use the same notebook to write your code and discussions.

Submission Details

• Work in groups of three and submit one lab per group.
• Deadline: April 9th at 23:59 (submission box below).
• The report will be graded—ensure all group members' names and SCIPER are included.
• Verify your code by rerunning the notebook before submission.
• Follow the naming convention: 
  lab_2_group_<group_id>.ipynb
• The notebook will be rerun with the provided libraries. No additional packages are allowed.

The Lab 3 is available on GitHub IAPR: https://github.com/LTS5/iapr2025
Use the same notebook to write your code and discussions. The routine(s) from the second part may be useful for the final project, so pay close attention.

Submission Details

• Work in groups of three and submit one lab per group.
• Deadline: April 16th at 23:59 (submission box below).
• The report will be graded—ensure all group members' names and SCIPER are included.
• Verify your code by rerunning the notebook before submission.
• Follow the naming convention: 
  lab_3_group_<group_id>.ipynb
• The notebook will be rerun with the provided libraries. No additional packages are allowed.

🍫 IAPR 2025 – Final Project Overview

The final project for IAPR 2025 will assess your skills in image analysis and pattern recognition. Your task is to implement a chocolate detector that can count chocolates of predefined types in an image.

---

🏆 Project Format: Kaggle Competition

This project will run as a Kaggle competition, where you will:

• Access training and testing data;

• Submit predictions and track your performance on a public leaderboard;

• Compete for fun!

⚠️ Note: The Kaggle public leaderboard is not part of the official grading, while the private leaderboard (released one day after the deadline) is: you'll have to beat the TA's baseline to get the points. Additionally, top-performing teams will receive bonus points and chocolate boxes! 🎁

You are free to use any techniques covered in the course—from classical image processing methods to advanced deep learning models. However, since deep learning approaches often yield significantly stronger results, and this is not primarily a deep learning course, we have created two separate Kaggle challenges to ensure a fair comparison across methods:

• Chocolate Recognition (Classic)
  ➤ Invitation Link: https://www.kaggle.com/t/316191fa57da43df8464e49cde94745b
  For solutions not using deep learning

• Chocolate Recognition (Deep Learning)
  ➤ Invitation Link: https://www.kaggle.com/t/fffe48b6588d4ffe86afae15ea498524
  For solutions using deep learning

🚨 You must choose one challenge to participate in. You may switch between challenges at any point during the project period.

👉 Read the rules of your selected challenge carefully to understand the allowed resources and constraints. 
👉 To access the links, please sign in to Kaggle using the "Sign in With Email" option and use your EPFL email address.

---

📦 Submission Requirements

Each group must submit a

archive (max size: 500MB) by May 21st at 11:59 PM, following the structure defined in the

project/README.md

on the IAPR 2025 GitHub Repository.

Your archive must include the following:

1. 📝 Jupyter Notebook Report

The report is a critical component of your project. It must focus on:

• Clear justification of design choices, explaining how each component in your method contributes to performance;

• Detailed technical descriptions of your solution;

• Strong quantitative and qualitative analysis, demonstrating and explaining the ability of your solution to segment the chocolates from the background and to compute relevant descriptors for each chocolate type, through means of metrics and visualizations (e.g. confusion matrix, class response heatmaps, clustering metrics, high-dimensional separability, learned filter visualization, etc.)

2. 🐍

main.py

(Executable Script)

This script should:

• Produce the exact submission file uploaded to Kaggle;

• Use only Python packages covered in Labs 1–3 (Python 3.9).

3. 📁

src/

Folder

Include all additional code and files used by

.
If using deep learning, you may also include your model checkpoints, ensuring they can be loaded correctly.

---

📤 Submission Instructions

• Work in groups of three – one submission per group

• 📅 Deadline: May 21st at 11:59 PM;

• 🧾 Include the names and SCIPERs of all group members in your report, as well as the challenge you chose to participate in;

• 🗂️ Use the following naming convention: 

  final_group_<group_id>_<kaggle_group_name>.zip<span style="color:rgb(51,51,51);">.</span>

---

🗣️ Final Presentation

Each group will deliver a 7-minute presentation followed by an 8-minute Q&A with the TAs. TAs will read your report beforehand and prepare specific questions based on your work. Your presentation should:

• Summarize your approach and design decisions;

• Highlight your main results;

• Be ready to answer questions on any part of your pipeline.

🗓️ Date: Friday, May 23rd
🕗 Time: Between 8:00 AM – 1:00 PM
📍 Location: In person (see detailed schedule below)

Schedule

Time	ELD 016	ELD 138	ELD 120	DIA 003
8h00	Group 22	Group 50	Group 44	Group 43
8h15	Group 14	Group 10	Group 9	Group 2
8h30	Group 37	Group 4	Group 52	Group 55
8h45	Group 75	Group 64	Group 46	Group 47
9h15	Group 23	Group 39	Group 1	Group 12
9h30	Group 5	Group 40	Group 34	Group 35
9h45	Group 33	Group 80	Group 74	Group 31
10h00	Group 13	Group 24	Group 26	Group 77
10h30	Group 7	Group 32	Group 38	Group 19
10h45	Group 69	Group 20	Group 48	Group 51
11h00	Group 15	Group 8	Group 73	Group 17
11h15	Group 18	Group 67	Group 25	Group 66
11h45	Group 21	Group 41	Group 56	-
12h00	Group 11	Group 3	Group 36	Group 27
12h15	Group 49	Group 42	Group 29	Group 30
12h30	Group 60	Group 28		Group 16

---

📚 Project Resources

• Classic Challenge – Kaggle - [Invitation Link]

• Deep Learning Challenge – Kaggle - [Invitation Link]

• IAPR 2025 GitHub Repository

• Project Slides

---

📊 Grading Scheme (100 Points)

🏅 Bonus Points (Kaggle Ranking)

🚫 Penalties

• Violation of Kaggle rules: –50 points