Tensor-flow_and_image

CS-63 Big Data Analytics Project

Goal:

Demonstrate the technology that can enable artists/painters to apply their imaginative styles from an image to any targeted image, generating a visual art from machine learning technique such as deep learning (hidden/deep convolutional neural network) and utilizing the massive available computing power (AWS GPUs)”

Visual Art-style transfer:

Demonstrate Style Transfer and Color transfer technique for Visual Art generation

Different Optimizer:

Compare performances of GradientDescent Optimizer, Adagrad Optimizer, Adadelta Optimizer, Adam Optimizer, RMSProp Optimizer and L-BFGS optimizer in visual art generation

Dataset and Model used:

Pre-trained Very deep Convolutional neural network Model of 16(VGG16) and 19(VGG19) layers is used. Dataset is a 82,000 Images taken from COCO Captioning Challenge( http://mscoco.org/dataset/#download)

Big Data Tools and Platforms/Technology used

1. Tensorflow

2. Deep Convolution Neural Network

3. Keras

4. Amazon EC2 instance with p2.8xlarge

5. Nvidia Tesla GPU K80

6. Filezilla

7. Flyod Hub

8. Anaconda Python

9. AMI-Udacity_dl AWS instance