Skip to content

Semantic Segmenter focused on masking high foliage areas within top down drone images.

License

Notifications You must be signed in to change notification settings

SequeI/BioMasker

Repository files navigation

Biomasker: Dense Foliage Prediction using PyTorch

Introduction

Biomasker is a novel deep learning model that utilizes a pyramid convolutional neural network (PyCNN) architecture to accurately predict dense foliage within aerial drone images. Developed using PyTorch, Biomasker achieves an impressive accuracy of 89.2%, demonstrating its effectiveness in extracting and classifying vegetation patterns.

Project Objectives

The primary objectives of this project were to:

  1. Design and implement a PyCNN architecture for dense foliage prediction from aerial drone images.
  2. Train and evaluate the PyCNN model on a comprehensive dataset of aerial imagery.
  3. Assess the model's performance in predicting dense foliage with high accuracy.

Project Methodology

  1. Data Preparation: The dataset consisted of aerial drone images with corresponding ground truth labels indicating the presence or absence of dense foliage. The images were preprocessed to ensure consistent dimensions and format.

  2. PyCNN Architecture: The PyCNN architecture employed a multi-scale feature extraction approach, combining downsampling and upsampling layers to capture both global and granular features. This allowed the model to effectively identify both coarse vegetation boundaries and finer details, such as individual trees or shrubs.

  3. Model Training: The PyCNN model was trained using the optimizer and the loss function. The training process involved optimizing the model's parameters to minimize the prediction loss and improve its overall accuracy.

  4. Model Evaluation: The trained PyCNN model was evaluated on a separate validation set to assess its performance in predicting dense foliage on unseen data. The model's accuracy was measured using the metrics of .

Project Results

The PyCNN model achieved an impressive accuracy of in predicting dense foliage from aerial drone images. This performance demonstrates the model's ability to accurately extract and classify vegetation patterns, making it a promising tool for various applications in land cover assessment, environmental monitoring, and natural resource management.

Project Conclusion

Biomasker has successfully demonstrated the effectiveness of PyCNN in dense foliage prediction from aerial drone images. The model's high accuracy and adaptability to diverse datasets suggest its potential for real-world applications in environmental monitoring, land cover analysis, and resource management.

Future Directions

Future research directions include exploring the use of Biomasker to monitor vegetation changes over time, incorporating additional data sources such as satellite imagery, and developing techniques for real-time foliage prediction.

About

Semantic Segmenter focused on masking high foliage areas within top down drone images.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages