rusty1s / deep-learning-on-graphs

Deep Learning

Resources

• Awesome Deep Vision
• 8 Inspirational Applications of Deep Learning
• Neural Networks and Deep Learning (Free Online Book)
• A Guide to Deep Learning
• Deep Learning Weekly (Newsletter)
• TensorFlow (Open Source Library)
• TensorFlow in a Nutshell
• A Year of Artificial Intelligence
• Introduction to Convolutional Neural Networks

Python Tutorials

• Learn Python the hard way
• Style Guide
• Principle component analysis in python with matplotlib

TensorFlow Tutorials

• Official
• Examples

Paper

• A method for stochastic optimization
  • Adam-Optimizer
• VGG 16/19
  • VGG in TensorFlow
• Kontinuierliche Filter
  • Deep Function Machines: Generalized Neural Networks for Topological Layer Expression (Section 4.1 und 4.2)
• CNNs auf Daten mit variierenden Größen
  • Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
  • A convolutional neraul network for modelling sentences
• Efficient and Scalable Computations with Sparse Tensors

Reduce Overfitting

• Dropout: A Simple Way to Prevent Neural Networks from Overfitting
• L2-Regularisierung
• Batch-Normalisierung

Data

• CIFAR-10 (Citation)
  • How to use
• ImageNet
  • 224x224 klassische Größe für allgemeine Bildklassifikation
  • ILSVRC2014
  • Random Ramblings on ImageNet
• PascalVOC

Applications

Object Detection

• Distance detection of objects
• Detect friends on images based on their facebook profiles
• Detect emotions from photos and represent them with an emoji
• Detect rotations/poses of objects and replace them with objects that have a similar rotation

Vector Graphics

• Convert svgs as graphs to input them in neural networks
• Convert pixelbased images to vectorbased images
• Fake Chine Characters in Vector Format with TensorFlow
• Create responsive svgs via Attention Detection (see SVG has more potential)
• Comparison: Performance and accuracy tests on different networking tasks like classification on pixelbased images and pixelbased images converted to vector graphics

Fonts

• Font Variant Generation
• Anaylizing 50k fonts using deep neural networks

Image Generation

• LowPoly Image Creation
• Create an Image from a Sketch
• Image Analogies
• Paint Photos in Style Of Any Artist
• Create an image from a text (A pictures says more than thousand words)

Graph-based image classification

• Efficiency and Performance by converting images to superpixels, segmentation graphs or svgs

deep-learning-on-graphs's Issues

Slides