Plant Identification Model with EZKL #149
-
Hi, I'm working on a project called WEFA where it incentivizes people to care for plants by earning digital rewards to play AR games. A key piece is the ML plant identification and health assessment to determine if the plant chosen is in the curated list and right zone along with health wise how much to reward a user. We're currently looking to use plant.id's API, ideally it would be nice to run the models on chain though we're not sure how feasible that is and how to source the models that would be compatible. We have a list of 16 plants with 4 in each category (herbs, fruits, flowers, vegetables) so potentially the model can be split into 4 pieces or even 1 for each plant. I don't have a ton of experience with models mainly using APIs for ML functionality but I'm willing and wanting to learn more with a dedicated use case. Would love any thoughts or feedback thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
Hey ! Probably best to create a single classifier across all categories (so 4*16 classes if I've understood correctly). Once you've built the model in pytorch you can run |
Beta Was this translation helpful? Give feedback.
Hey ! Probably best to create a single classifier across all categories (so 4*16 classes if I've understood correctly). Once you've built the model in pytorch you can run
ezkl
on it to create a model that can submit proofs on-chain + a smart contract that can verify said proofs :)