We employ various approaches and algorithms in the training of our models, all falling within the domain of one-shot learning. The crux of this approach lies in the initial phase where the neural network learns to differentiate between objects of different classes. Subsequently, in the second stage, it hones its ability to determine the most suitable class for the input image. This two-stage process yields favorable results, especially in scenarios with limited training samples.
The disease and crop models, integrated into the platform a few weeks ago, exhibit a statistical accuracy exceeding 99%. Upon analyzing user requests, it is evident that these models outperform their predecessors in real-world conditions.
As always, for plant care tips and to check on your plants' health, utilize our DoctorP application. It's available on Google Play, the App Store, or through our Telegram bot at https://t.me/PlantsDoctor_bot
