bitisony / fastai_deep_learning_course_part1_v3

My sideprojects for fast.ai's Practical Deep Learning for Coders course (fast.ai part1v3)

Fast.ai Practical Deep Learning for Coders (Fall 2018)

Part I of Fast.ai's two-part deep learning course, offered through The Data Institute at USF. From October through December of 2018. Part II is here.

A top-down approach to becoming a deep learning practitioner. We first learned how to obtain state-of-the-art results in several deep learning tasks. Then we dug deeper, developing an intricate understanding of the theory behind the techniques we apply.

My Mini Sideprojects

• Blue Jay Species Classifier
  • Writeup on Medium
• Implementing Momentum from Scratch
• Writing My Own Version of PyTorch's nn.Linear Class

My Reimplementations of Lesson Notebooks

Week 1: Intro to Computer Vision

• Recognizing Cat and Dog Breeds

Week 2: Computer Vision - Deeper Applications

• Gradient Descent from Scratch

Week 3: Multi-Label Segmentation, Image Regression, and More...

• CamVid Segmentation SOTA
• Biwi Kinect Head Pose Image Regression
• Kaggle Planets Competition Multi-Label Classification
• IMDB Sentiment Analysis SOTA

Week 4: NLP, Tabular, and Collaborative Filtering

• MovieLens Collaborative Filtering
• Neural Nets for Tabular Data

Week 5: Foundations of Neural Networks

• Stochastic Gradient Descent on MNIST

Week 6: Regularization and CNNs

• Weight Embeddings for Categorical Features in Tabular Data
• Convolutional Kernels and Coding Heatmaps from Scratch

Week 7: ResNets, UNets, GANs, and RNNs

• Understanding ResNet Blocks
• Super Resolution using UNet and Perceptual Loss (on Pet Images)
• Super Resolution using a GAN with Pre-Trained Generator and Discriminator Networks
• Training a WGAN to Generate Life-like Images from Noise
• Demystifying RNNs, GRUs, and LSTM Cells

USF Completion Certificate

Verify at Open Badge

Dependencies

• requirements.txt
• environment.yml