Model card
Structure-based permeability model
What the model predicts, how it was tested, and exactly where the product uses it.
The machine-learning model predicts a polymer's oxygen and carbon-dioxide permeability from its chemical structure. It was trained on published measurements for several hundred polymers and tested on polymers it never saw, where it clearly beats a baseline that ignores the chemistry. The product uses it to screen films that are not in the database and to cross-check sourced values, always labelled as a model prediction with its range. It never overrides a sourced value and never chooses the packaging.
Screen a film that is not in the database
Pick a polymer or paste its repeat unit (SMILES with two * atoms). The answer is a model estimate for screening, never a recommendation.
The model runs on the project's own machine and cannot be reached right now. Everything else on the site works.