The source code for our team's submission to IC Hack 23. A web app that can scan an image of a product, identify brand logos, and look up the brands in a database to return 'ethics scores'. The aim is to help consumers make more conscientious decisions when shopping. Uses Flask for the backend, Keras for computer vision (brand logo recognition), and other libraries such as Pandas and Beautiful Soup for web scraping to generate ethics scores.
Three data sources:
- Data source 1:
- https://www.worldbenchmarkingalliance.org/publication/chrb/rankings/type/ungp/industry-apparel/
- Webscrape the above data (also click onto each site and record the scores in the different metric such as "remedies and griance mechanisms"), e.g. https://www.worldbenchmarkingalliance.org/publication/chrb/2020/companies/tesco-2/
- Data source 2:
- https://drive.google.com/file/d/1Nc75LmR2oKUsIwvP5ycG6ZbpPvuowJDR/view
- Put this above in a CSV
- Find a way of aggregating this to produce a transparency score
- Data source 3:
- https://www.cdp.net/en/responses?queries%5Bname%5D=nestle
- Does a dynamic search on this website based on input brand name
- Looks up the climate change report and returns the letter score
Run the following script to download the weights for the brand logo recogniser brand_recogniser_model/build/build.sh.
Install necessary python modules (or use pip):
conda install matplotlib keras=2.2.4 pillow scikit-learn
conda install tensorflow-gpu
conda install opencv=3.4.4`