DevMesh

Automated Weed Detection and Removal in Agricultural Fields Using oneAPI Toolkit

Nitin Mane

Nitin Mane

Aurangabad, Maharashtra

0 0
  • 0 Collaborators

The proposed method uses a Residual Network (ResNet) architecture with Auto Mixed Precision (AMP) to classify images of crops and weeds. The use of AMP allows for dynamic adjustment of the precision of computations during training and inference, which can improve the model's performance. ...learn more

Project status: Under Development

oneAPI, Robotics, HPC, Artificial Intelligence

Intel Technologies
DevCloud, oneAPI, Intel Python, Intel vTune, MKL, Intel CPU, Intel GPA

Code Samples [1]

Overview / Usage

The objective of this project is to develop a computer vision model that can accurately detect and remove weeds in agricultural fields with minimal environmental impact. The proposed approach uses a Residual Network (ResNet) architecture with Auto Mixed Precision (AMP) to classify images of crops and weeds. The use of AMP allows for dynamic adjustment of the precision of computations during training and inference, which can improve the performance of the model. The use of oneAPI Analytical Toolkit will enhance the performance and efficiency of the model, leading to improved crop yields and more efficient use of resources.

Methodology / Approach

The implementation of this project will involve the following steps:

  1. Download a dataset of images of crops and weeds from real-world environments.
  2. Use the oneAPI Data Analytics Library (DAL) to preprocess the dataset and prepare it for training.
  3. Train a ResNet with AMP to classify images as either crops or weeds.
  4. Optimize the model's performance using the oneAPI Math Kernel Library (MKL) and oneAPI Graph Analytics Library (GAL)
  5. Use the oneAPI Model Analyzer and Debugger (MAD) to analyze the model's performance and identify any areas for improvement.
  6. Deploy the trained model to a simulated production environment using the oneAPI Deployment Manager.
  7. Use the oneAPI Performance Profiler to monitor the model's runtime performance and identify any bottlenecks that may be impacting its performance.
  8. Continuously monitor the model's performance and make updates as needed to improve its accuracy.

Repository

https://github.com/Nitin-Mane/Automated-Weed-Detection

Comments (0)