The problem KhetiBaadi solves

It is an offline Android and Web application that helps farmers predict disease in their crops so that they can take apt actions. It successfully detects plant disease using leaf picture with high accuracy (97%), detects around top 3 disease probabilities to help users among other chances, provides solutions for the disease detected. The platform is made available offline so that it can be accessed anytime and anywhere. It classifies non-leaf images as ‘background’ images so that the model doesn’t give random answers.

Challenges I ran into

The main challenge was faced while converting trained PyTorch model to tflite for deployment. The solution was to convert it via Keras model as an intermediate.