/analyzing_56k_fonts_using_deep_neural_network

Final project for E6893 Big Data Analytics

Primary LanguageJupyter Notebook

Final project for E6893 Big Data Analytics

Group Index: 201605-96

Group member:

  • Ruixiong Shi rs3569
  • Yuchen Shi ys2901

Training Network:

  • model.py: model strucutre written and auxiliary function
  • train.ipynb: Notebook for training the network
  • model.pickle.gz: Pickled file network parameters

Visualization

  • pyspark_visual.ipynb: Visualize dataset by spark
  • read.py: auxiliary function for reading hdf5 file into spark rdd file
  • TSNE2D.ipynb: Visualize embedding space in 2D with T-SNE
  • tf-embedding-visualization: Visualize embedding space in 3D with T-SNE

Generating new fonts

  • embedding_perturb_multinormal.ipynb: Creating new fonts by random perturbation and sampling from multivariate normal distribution of original embedding space

  • random_perturb: output character created by random perturbation

  • multivariate_norm: output character created by sampling from multivariate normal distribution

  • interpolation.ipynb: Creating new fonts by interpolating embedding space

Download Dataset

$ wget https://s3.amazonaws.com/erikbern/fonts.hdf5  

Youtube Video

Please click here