Case study card:
Computer vision to identify brands responsible for waste
There is talk among legislators about putting the burden/cost of litter collection on the brands found in litter.
The challenge was to identify both the brands of packaging found and the material weights for recycling purposes using computer vision to automate identification.
Further identification of brands to categorise waste.
Identifying brands and products
Using Computer Vision and OCR
Trained on very large data set
High speed performance for inference
Runs on Jetson (Nvidia)
Model is custom and highly optimized
Achieved very high accuracy F1 score of ~98%
Confidential (recycling body)