<description>This week I was busy applying for jobs and working on the CS38220 assignment. So I only managed to do code clean-up and implementing a config file which allows a user to change the order of filters and other settings.
Config file example: config.yml
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 --- # Config file for autophotographer # List of filters to apply in order filters: - brightness - filesize - contrast - focus # Whether or not to apply CNN ranking CNNrank: True # Options for focus filter brightness_options: threshold: 0.</description>
<pubDate>Sun, 27 Mar 2022 13:13:48 +0100</pubDate>
<guid>https://mmp.oscar.blue/posts/week-8/</guid>
<description>Monday # Towards the end of last week ( Week 7), I managed to refactor my code in order to make it more portable. This allowed me to train my model on different machines. I ran my training script on the uni&rsquo;s GPU compute successfully for 20 epochs. The next stage was to train it for longer and analyse the results. On the Monday morning I adjusted the parameters of my training script to train for 2000 epochs instead.</description>
</item>
<item>
<title>Week 7</title>
<link>https://mmp.oscar.blue/posts/week-7/</link>
<pubDate>Sun, 20 Mar 2022 12:40:18 +0100</pubDate>
<guid>https://mmp.oscar.blue/posts/week-7/</guid>
<description>Now that I had successfully run my model without any runtime errors, the next step this week was finding some GPU compute so I can train my model on much more powerful hardware to accelerate the training.
My first idea was to use cloud computing. There are machine learning specific cloud technologies, but I didn&rsquo;t want to use these as I didn&rsquo;t want my code to be dependent on the specific ways cloud platforms want the code in.</description>
</item>
<item>
<title>Week 6</title>
<link>https://mmp.oscar.blue/posts/week-6/</link>
<pubDate>Sun, 13 Mar 2022 12:40:17 +0100</pubDate>
<guid>https://mmp.oscar.blue/posts/week-6/</guid>
<description>This week I finished programming the basic CNN model using transfer learning. I decided to train it for 20 epochs to make sure there weren&rsquo;t any runtime errors in my code. As I don&rsquo;t own an Nvidia GPU (I have an AMD GPU), I couldn&rsquo;t make use of the pytorch version that utilised CUDA to speed up processing. There is a RocM version of pytorch for AMD GPUs1 but RocM isn&rsquo;t as mature as CUDA and only officially supports a small subset of Linux distributions.</description>
</item>
<item>
<title>Week 5</title>
<link>https://mmp.oscar.blue/posts/week-5/</link>
<pubDate>Sun, 06 Mar 2022 12:40:14 +0100</pubDate>
<guid>https://mmp.oscar.blue/posts/week-5/</guid>
<description>Starting to write a CNN # This week I started to implement what I had learnt about CNNs in Week 4. At this point I hadn&rsquo;t designed a CNN architecture to implement, instead I wanted to have a running model regardless of performance just to see if I could implement one and understand it.
Half way through implementation, I decided to look back at the existing research papers on judging aesthetic judgement to see which aspects of their system and CNN were important to the task.</description>
<description>This week I did some research into how to build a CNN from scratch, including the different type of layers, loss functions, learning rates, epochs and other core concepts.12
I also set up and created this blog with Hugo to document my progress and setup Woodpecker CI to do continuous testing and integration.
<pubDate>Sun, 20 Feb 2022 12:46:55 +0000</pubDate>
<guid>https://mmp.oscar.blue/posts/week-3/</guid>
<description>Filming footage # At the start of the week I went into town to film some practise footage to work with later (up until this point I had been experimenting with footage limited by my bedroom walls). I took some basic vertical and horizontal footage of the town - no nature or breach footage yet.
Gaining more useful information # I used the footage I had recorded at the start of the week and revised my &ldquo;filesize&rdquo; code.</description>
</item>
<item>
<title>Week 2</title>
<link>https://mmp.oscar.blue/posts/week-2/</link>
<pubDate>Sun, 13 Feb 2022 12:46:54 +0000</pubDate>
<guid>https://mmp.oscar.blue/posts/week-2/</guid>
<description>This week I set up my repositories and starting writing some basic code.
Set up # Before starting any coding, I wanted to set up my remote git repositories. I had already decided I wanted the project mirrored over two remote git repositories from different providers as a safety precaution. My intial plan was to use the university&rsquo;s GitLab instance but as it&rsquo;s recently been moved behind the firewall it would have made mirroring quite difficult.</description>
</item>
<item>
<title>Week 1</title>
<link>https://mmp.oscar.blue/posts/week-1/</link>
<pubDate>Sun, 06 Feb 2022 12:46:51 +0000</pubDate>
<guid>https://mmp.oscar.blue/posts/week-1/</guid>
<description>This week is the first week of the project. I researched academic papers, existing code and dataset relating to the topic of determining aesthetics.
Papers # Photo Aesthetics Analysis via DCNN Feature Encoding1 - Predicting aesthetic performance using a bespoke CNN solution
AVA: A large-scale database for aesthetic visual analysis2 - Making of an aestehtic visual analysis dataset
Code # Image Quality Assessment - Convolutional Neural Networks to predict the aesthetic and technical quality of images.</description>
