machine_learning_project

A group project for cmpt 419 at sfu focused on exploring the growing area of machien learning

Group Members:

Kier Lindsay

Rafid Ashab Pranto

Vinaik Chhetri

Concept

Using a unet to improve detale in style tranfer

Keras implementatations of unet and style transfers

https://github.com/robertomest/neural-style-keras

https://github.com/zhixuhao/unet/blob/master/model.py

Data Sets

http://cocodataset.org/#download

https://bam-dataset.org/#explore

https://artuk.org/discover/artworks/view_as/grid/search/has_image:on--class_title:landscape/page/2#

Other links

A blog on style transfer https://shafeentejani.github.io/2017-01-03/fast-style-transfer/

another keras implementations https://github.com/misgod/fast-neural-style-keras

We Will be Looking at testing the usage of pose transformation and pose extraction out of pictures. with the help of Vunet: https://github.com/CompVis/vunet https://arxiv.org/pdf/1505.04597.pdf //the cited paper on unets used in vunet

https://ml5js.org/docs/style-transfer-image-example //Idea for somting to do with vunet have it transfrom a picture you give it to a pose in the browser or in a app. Even better if you can edit the pose live.

https://github.com/tensorflow/tfjs-models/tree/master/posenet //an example of live pose detection it does not do transformations though

References

• [1]: L. A. Gatys, A. S. Ecker and M. Bethge. "A Neural Algorithm for Artistic Style". Arxiv.
• [2]: J. Johnson, A. Alahi and L. Fei-Fei. "Perceptual Losses for Real-Time Style Transfer and Super-Resolution". Paper Github
• [3]: V. Dumoulin, J. Shlens and M. Kudlur. "A Learned Representation for Artistic Style". Arxiv Github
• [4]: D. Ulyanov, A. Vedaldi and V. Lempitsky. "Instance Normalization: The Missing Ingredient for Fast Stylization". Arxiv
• [5]: Olaf Ronneberger, Philipp Fischer, and Thomas Brox. "U-Net: Convolutional Networks for Biomedical Image Segmentation". Arxiv
• [6]: Patrick Esser, Ekaterina Sutter, Bj ̈orn Ommer. "A Variational U-Net for Conditional Appearance and Shape Generation". Paper Github
• [7]: T. Lin et al. "Microsoft COCO: Common Objects in Context". Arxiv Website