Training neural networks to reason about album covers and predict mood/tone of the album
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
embedding
output
proposals
refData
sandbox_extras
.gitignore
1a_parse_weeks_script.py
1b_parse_reviews_script.py
2a_pickle_switcher.py
2b_resize_images.py
3_exploratory_analysis.ipynb
4a_build_models_basic_basicmeta.ipynb
5_human_ranking.ipynb
5_predict_and_evaluate.ipynb
README.md
helper_functions.py
train_embedding_models.py

README.md

ALBUM NET

By: Brady Fowler & Seth Green

  • Training neural networks to reason about album covers and predict mood/tone of the album.

BLOG

Data Collection

Use BeautifulSoup package in Python to scrape AllMusic.com for covers, “moods,” and reviews. No statistics have been found that suggest the total number of covers/reviews hosted by Allmusic but we hope to have at least 50,000 training examples from a variety of time periods and genres.

The Directory Structure is:

┳━ album net/
┗━┳━ data/
  ┗━┳━ images/ * all raw album covers *
    ┣━ links/  * one pickle per album *
    ┣━ moods/  * one pickle per week  *
    ┗━ sized_images/ * all sized album covers *