AI image Processing - artistic outputs
Budget: $30 – $250 USD
Looking for an expert in AI image processing such as StyleGan, StyleGAN2, Pix2Pix and more. We are already familiar with the technologies and can generate great picture qualities and we are familiar with most algorithms such as the ones below. We need someone to have deep knowledge of the technologies and able to create new technics/algorithm to be able to generate even greater quality of pictures. We want to transform a picture to an artistic paint using different filters. For example, we need to be able to transform a picture into a watercolor paint. Or into a oil paint, etc..
1) Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization
https://github.com/xunhuang1995/AdaIN-style
https://arxiv.org/abs/1703.06868
2) Perceptual Losses for Real - Time Style Transfer and Super - Resolution
https://github.com/jcjohnson/fast-neural-style
https://arxiv.org/abs/1603.08155
The algorithm we need to develop must have the following properties:
a) Be able to train a filter, like watercolors arts, oil painting arts and custom arts
b) Be able to indicate that, when transforming a picture,
i. we can keep the original colors or the filter colors.
ii. Able to increase/decrease size of detailing (paint brush strokes, line marks, etc..)
iii. Respect the original picture size. The output size matches the input.
c) After the training is done and a new filter is generated, the output picture generated will take seconds. Very similar to Fast-Neural-Style.
1) Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization
https://github.com/xunhuang1995/AdaIN-style
https://arxiv.org/abs/1703.06868
2) Perceptual Losses for Real - Time Style Transfer and Super - Resolution
https://github.com/jcjohnson/fast-neural-style
https://arxiv.org/abs/1603.08155
The algorithm we need to develop must have the following properties:
a) Be able to train a filter, like watercolors arts, oil painting arts and custom arts
b) Be able to indicate that, when transforming a picture,
i. we can keep the original colors or the filter colors.
ii. Able to increase/decrease size of detailing (paint brush strokes, line marks, etc..)
iii. Respect the original picture size. The output size matches the input.
c) After the training is done and a new filter is generated, the output picture generated will take seconds. Very similar to Fast-Neural-Style.
Related categories:
Algorithm
Machine Learning (ML)
Artificial Intelligence
Neural Networks
Deep Learning