Muhammad Abdul Rehman Shah

Academic

Career

Verbatim text

Pest Recognition of Images using Deep Learning 
 
Our project focuses on using various deep learning techniques on images of a variety of pests to 
come up with an effective model to classify images of pest. This project has direct significance in 
the field of agriculture where tones of crops are wasted just because farmers can’t recognize and 
prevent pest attacks on their crops. According to research (At Least 40% Global Crops Lost to Pests 
Every Year: FAO, n.d.) 40% of the global crops are annually wasted to pests amounting to around 
€69 billions of economic loss that is enough to feed around 1 billion people annually.  
We discovered a dataset of labelled pest images called IP102 (Wu et al., 2019) which comprises 
more than 75000 images of 102 classes of pests. 
During FYP-1, we focused on developing a model that classifies the images of pests in the IP102 
dataset and used different feature extraction, data pre- processing and deep learning techniques to 
prepare our dataset and train it to come up with a base line model. Then we fine-tuned our model 
and performed data augmentation techniques to increase its accuracy in FYP-2. We then embedded 
the best model as a TFLite model into an iOS mobile application using Swift. 
 
 
 
 
 
 
 
Technology Used: 
Python, TensorFlow, TFLite, OpenCV, 
 PIL, Swift, Kaggle 
Supervisor Name: 
Dr Labiba Fahad 
Group Members:   
Muhammad Abdullah (I18-0416) 
Muhammad Abdul Rehman Shah (I18-1566)

AI enrichment