Home

Author

Affiliation

Jeremy Fix

CentraleSupelec

Keywords

Deep learning, practicals

Deep learning lectures © 2018 by Jeremy Fix is licensed under CC BY-NC-SA 4.0

Forewords These pages are written in markdown format and generated with quarto. The generation is performed using the github workflows. Thank you all for these very helpful tools.

These pages contain the subjects and codes for the labs in the deep learning lectures 3MD4040 taking place at CentraleSupélec. Most the work is done in pytorch although you will find one lab using Keras.

Labs

PyTorch Labs

FashionMNIST Classification

First steps in PyTorch: classifying fashion objects with neural networks

Object Detection

Transfer learning and object detection on Pascal VOC dataset

Semantic Segmentation

UNet and semantic segmentation on Pascal VOC with dense pixel labeling

Speech Recognition

Automatic speech recognition on Mozilla Common Voice using CTC models

Generative Networks (GAN)

Deep convolutional generative adversarial networks for image generation

Neural Implicit Representations

Using Neural Networks for implicitly encoding data : images and spatio temporal cardiac MRI

Keras Labs

MNIST Classification

First steps in Keras: classifying handwritten digits with various architectures

🎤

CIFAR-100 Classification

Additional ressources

Practicals/Lectures

• Oxford’s CNN practicals
• Nando de Freitas lectures and practicals
• Stanford A. Karpathy practicals
• Master Data Science Paris Saclay
• Fast.ai, J. Howard

Blog posts

• An overview of gradient descent algorithms
• Revisiting sequence to sequence learning, with focus on implementation details
• The unreasonnable effectivness of recurrent neural networks
• Theories of Deep learning