Project Goal - To propose a low-cost solution for weed detection in Sugar Beet plants by solely using an RGB camera.


Solution -

  • Implemented a machine learning model for Pixel Wise Semantic Segmentation of weed, crop, and background.
  • The model is a CNN based encoder-decoder network with residual blocks.
  • The input is an RGB image which is stacked with 11 more channels like vegetation indexes, HSV color space, operators on the indices such as the Sobel derivatives, the Laplacian, and the Canny edge detector.
  • The 14 channel input helps in the training and generalization of the model due to a small dataset and a wide variety of environmental conditions.
  • The output is an RGB segmented image of crop, weed, and background Python Implementation - Keras, TensorFlow, OpenCV.
  • Used my local GPU Nvidia GTX 960m and Google Colab's free GPU (Tesla K80) for training & testing purpose.


Github URL -



You Might Also Like


Django, Java, Keras-Tensorflow, PyTorch

Weed Detection in Sugar Beet Plants

Keras-Tensorflow, OpenCV, Pixel Wise Semantic Segmentation

Blogging Platform

CSS, HTML, JavaScript, jQuery, JSP, MySQL, Servlet

Copyright 2024 © Saurabh Bora