Image color topic modeling using fastTopics.
The original idea is from my final project for the course HG48600: Fundamentals of Computational Biology: Models and Inference (poster).
This project uses renv to ensure reproducibility. The dependency information is store in renv.lock.
Run renv::restore() to restore the exact version of dependencies
from renv.lock. This will create a project-specific library under
renv/ and create an .Rprofile to use that library when the project
is opened.
R/: R scripts used for the project. Files are named with numerical prefixes, indicating the order in which they should be executed for successful run-through.output/: Saves model snapshots for future reference or reuse. Also, important intermediate results generated during data preprocessing and model training and stored here.images/: All graphical outputs generated during the analysis, primarily aimed at visualizing the model's performance and characteristics, are stored here.data/: Stores the input dataset used in the project. See below on the example data.
To run the analysis on the example data, first use
git lfs install
git clone https://github.com/nanxstats/ChromaClust.git
git clone https://huggingface.co/datasets/nanxstats/movie-poster-5k
unzip movie-poster-5k/data/movie-poster-5k.zip -d ChromaClust/data/A minimal Shiny app is built for easy review of images under
the same color topic or mixture of topics.
To use the app, open the project and run through R/5-exemplar.R.
The example below shows a pattern from the combination of three color topics:
- Dark color background (topic 7).
- High contrast color title (topic 9).
- Protagonist with skin color shown in the foreground (topic 10).
