m-prth / cotton_diseases_detection

Cotton Diseases Detection: Transfer Learning

By Aayush Jain(Darkshadow)

• Created a Deep Learning Pipeline which will use pre-trained models(transfer learning) for making predctions on cotton diseased plants.
• The datasets consists of images of cotton plants with diseases.

---

Code and Resources used:

Python: 3.7
Libraries Used: Tensorflow, numpy, glob Tensorflow Version: 2.2.0 Link of Dataset: Dataset

---

Model Description:

• We have used two pre-built models called Resnet15 and InceptionV3.
• They are available in Tensorflow keras api and also in the link given below.
• To get Resnet152V2 trained model : RESNET15V2
• To get InceptionV3 trained model : INCEPTIONV3

---

Accuracy and Loss:

---

Model Accuracy

• After training the data, watching the training accuracy and validation accuracy we can say that both the models are having almost same accuracy that's training accuracy ~98% and validation accuracy ~96%.

---

Production

• For production we can use InceptionV3 as it has less trainable parameters and we get similar accuracy as Resnet152V2 which as comparitively much high trainable parameters.

---