Crop Disease Detector

• Trained resnet-50 with one-cycle-policy for plant disease finegrained classification using fastai.
• The model is capable of classifying a total of 14 crop classes and beated the previous models with an accuracy of 99.73% within 1.6 hr of training on a Tesla P4 instance.
• The custom dataset is consisted of 38 classes from PlantVillage dataset and 1 background class from Stanford's open dataset of background images - DAGS, finally this was divided into a train-validation ratio of 80:20

Demo Code

Dog Breed Classifier

• Finetuned the resnet50 model (pretrained on IMAGENET) to identify the dog-breed in dog-images (out of 120 breeds)
• Used fastai for augmentation, training, validation and testing
• Owned a validation accuracy of 92.22%

Code