Blogpost #7
This weeks readings focused on the concept of montage and how the sequencing of images in time can be used stylistically to highlight certain aspects of the images contained within. In the case of Battleship Potemkin (and the Soviet School in general), the focus of montage is often on showing opposition (sailors vs. officers, one vs. many, etc.) and with an emphasis on dialect and logical/concrete meaning rather than spiritual or psychic composition.
For my semester project I have been focusing on learning how to use DeepDream (a gradient ascent method invented by google in 2015 using TensorFlow to find patterns in noise based on training data). I have fed it some previously-created images, shown here:
Without resampling image, the result is low resolution and blurry. I also attempted to find patterns in too small of regions, making the result indistinguishable from noise in some parts of the image (image 1). Resampling image to process smaller chunks + setting octaves (minimum size to find a pattern within) gives a better result (image 2). This was made using a pre-trained model from Keras called InceptionV3 (this is the default model used by DeepDream). I would like to try to run the same algorithm using different pre-trained models from both Keras and Caffe.
I have also been looking into Live Coding (a technique used to create music and images in real time using code). I attended a live performance at Zuzu last week, and have begun to learn Gibber (https://gibber.cc/). I hope to further explore generative art with a temporal aspect in both sound and image creation. This, by-happenstance, fits in nicely to the topic of the week on montage, since this technique is used to generate montages of both sound and image. I would be interested in trying to make Live Coding pieces inspired by the schools of montage in this week's reading. This week I will also be meeting with Prof. Nick Montfort (who performs Live Coding and researches generative art at MIT) this week to explore further and to discuss the IAP course I am creating.











