Uni, Institut d'Optique Graduate School, 2020
(with François Goudail, Stéphane Herbin, Adrien Chan Hon Tong, Alexandre Boulch.).
2020 Material
Description in the IOGS Course Panel - SynapseS
Note: Colab notebooks can be saved in your own environment using the “Copy to drive” item in the “File” menu.
| Date | Instructor | Topic | Course | Exercises |
|---|---|---|---|---|
| 14/01 | SH | Introduction to Machine Learning | course #1 | — |
| 20/01 | BLS | Decision trees, random forests and boosting | Course #2 | ipynb / colab / smile ref data for Olivetti faces / ipynb results / colab results |
| 21/01 | ACHT | Neural Networks | course #3 | — |
| 27/01 | SH | Support Vector Machines | course #4 | — |
| 28/01 | BLS | Dimensionality reduction and clustering | Course #5 | ipynb / colab / colab results |
| 03/02 | ACHT | Deep Learning | Course #6 | — |
| 10/02 | SH | — | Exam (PCA ipynb / html ) | mini-project starts: cell segmentation / adversarial attacks / dehazing (see below) |
| 11/02 | SH | Regression | Course #8 | mini-project |
| 18/02 | ACHT | Deep learning applications | Course #10 | mini-project |
| 25/02 | BLS | Generative Networks and Auto-encoders | Course #9 | mini-project ends |
Dehazing mini project
The Dehazing mini project span over 4 exercise sessions. It is based on the NTIRE 2020 challenge on Non homogeneous dehazing. The goal is to dehaze some images, i.e. removing haze, fog, mist and other smoke.
On the codalab page of the challenge, one can register and get access to the data. Data consists of a training set of 45 pairs of hazy and clean images, and a validation set of 5 hazy images.
We provide jupyter notebooks:
- One to parse the images and save them in numpy arrays: im2npy.ipynb and html
- A DataLoader for subsequent machine learning processing in pytorch: dehazing.ipynb and html (credit: A. Boulch)
Prepared numpy arrays can be downloaded: Train data / Train GT / Validation data. Please copy data locally, or put them in your own drive for use on Colab.
2019 Material
Description in the IOGS Course Panel - SynapseS
Colab notebook can be saved in you own environment using the “Copy to drive” item in the “File” menu.
| Instructor | Topic | Course | Exercises |
|---|---|---|---|
| SH | Introduction to Machine Learning | course #1 | — |
| SH | Support Vector Machines | course #2 | — |
| BLS | Decision trees, random forests and boosting | Course #3 | ipynb / colab / smile ref data for Olivetti faces / ipynb results / colab results |
| AB | Neural Networks | course #4 | — |
| BLS | Dimensionality reduction and clustering | Course #5 | ipynb / colab / colab results |
| AB | Deep Learning | Course #6 | — |
| SH | — | Exam | mini-project starts |
| SH | Regression | Course #8 | mini-project |
| BLS | Generative Networks and Auto-encoders | Course #9 | mini-project: data-loaders available |
| AB | Recurrent Neural Networks | Course #10 | mini-project ends |