This a challenge sponsored by Microsoft that was hosted on zindi platform
The objective of this challenge was to predict the disease of a rice plant giving it's RGB and Infrared images.
There are three possible categories for a plant:
- rice blast disease
- brown spot disease
- healthy
Variant types of (CNN-Transformer) based models were trained with variant sizes (ResNet-EffNet-ConvNext-Swin-Deit-Coat).
Simple resnet_152 achieves ~ 96.5% accuracy.