In recent times, there is high demand to design simple and low-cost IoT-based solutions for real-time testing of the paddy leaf pest detection and control system. It is a difficult task to design the real-time video streaming system using a local IP address and camera interfacing. This paper aims to test and evaluate the performance of a machine learning-based IoT-based feature extraction algorithm for such images. The paper contributes in two parts. In the first part, the challenge of testing the real-time IoT-based image acquisition system using ESP32 cam module is successfully addressed. The captured images and a database of healthy and damaged images are processed in the HSV space, and thresholding is applied to segment the desired brown spots. In the second part, the feature extraction algorithm is designed using a sequential image processing procedure. A rich set of statistical and contrast features are calculated for different healthy and gray spot paddy leaf images. Support vector machine (SVM)-based classifiers are tested and evaluated using the proposed feature sets.