The Litterati Open Data Portal allows anyone to access the raw data created by the Litterati community. This information includes:
Raw litter labels(A string of user-generated or LitterAI confirmed labels)
In your account, you will be able to visualize your data in two formats. First, you can see your data aligned to the Open Data Portal (see above) and the other is Litterati Filtered view. In Litterati Filtered view, you will see a combination of Litterati filtering and LitterAI tags (Litterati´s proprietary image recognition model to identify Object, Material, and Brand).
How Litterati Litterati Filtered View works:
Litterati will filter the raw litter label into a standard format of object, material, and brand. For example, if the user adds “butt” and “paper” to the litter photo, Litterati will filter the user input to “cigarettebutt” and “cellulose acetate.” We do this in order to improve the standardization of the naming conventions since many times user labels can contain spelling mistakes, spaces, multiple languages, and synonyms or aliases of the same item. We constantly update our filtering system to improve its performance and is based on the community experience we have developed over the last several years.
LitterAI is a collection of multiple image recognition models focused on identifying object, material, and brand of all litter. We process all uploaded images through our latest models to identity what is in the litter images. Then if we determine that the precision and accuracy is of a high enough confidence interval then we will apply the output of the model to the litter item.
How that becomes the Litterati Filtered data:
Is the combination of the above two processes to achieve the best Litter data possible, which allows for the best determination of what is in the litter image.
The Litterati Filtered data is post-processed to determine other relationships. The category can be related to a behavior or a sector to make it easier to determine how to tackle the litter issues. For example, “cigarettebutt” & “cigarettepack” are related to the smoking category. Whereas “straw,” “cup,” “lid,” “can,” and “bottle” are all mapped to the drink category. By categorizing litter, we can determine which actions should be taken to tackle the specific problem since actions on smoking litter are very different from drink-related items.
Litterati currently categorizes items by:
Understanding the relationship between litter’s weight, volume, and count is critical to understand the severity and size of the litter problem. From the Litterati Filtered output, major items such as bottles, cans, and cigarette butts are matched to standardized weight and volume to provide a more accurate picture of the litter issue.
There are over 50 waste taxonomies across the globe and everyone looks at taxonomies with a slightly different lens. With our granular taxonomy, we have been able to map to several of the major standards and we will add more in the future. Currently, we use a combination of the Convention for the Protection of the Marine Environment of the North-East Atlantic (OSPAR), Commonwealth Scientific and Industrial Research Organisation (CSIRO); Break Free From Plastic (BFFP), and National Oceanic Atmospheric Administration (NOAA) labeling taxonomies.
Brand Parent Corporations:
With thousands of global brands and hundreds of companies responsible for producing them, it is important to match these individual brands to their parent companies. For example: Fanta, Monster, Dasani, Fernandes, Aquarius, and Maaza, etc. are all owned by the CocaCola Corporation. Litterati is able to match brands to their corporate parents at scale. Currently, we have matched 500 brands to their respective parent companies and we will continue to add more in the future.