This project is about detecting Dengue’s vectors breeding site from Google Street View images using deep learning. The main vector is Aedes Egypti mosquito.

• Pipeline of process
• Description of Code
• Directory Structure
• Getting Started
  • Prerequisites
  • Instalation
  • Running
  • Evaluating model accuracy
• Demo
• Built With

• data_collection.py used for retrieve google street view images of village that you want. The image size is 600x600 pixels.

• image_recognition.py used for recognize images , it will return top five classification results.

• image_segmentation.py used for extract segmented of image.

• feature_vector_classification.py used for classify the result of image recognition and image segmentation again for increasing more accuracy.

+Mosquito_Breeding_Sites_Detector
  +SegNet-Tutorial
  +caffe-segnet
  +GSV
  +xgboost
  +dataset
  +geojson
  +scripts
    +feature_vector
      -to_geojson.py
      -xgb_classifier.py
    +image_processing
      +model
      -image_divider.py
      -image_recognizer.py
      -inception.py
    +image_retreival
      -GSV_loader.py
      -get_village_points.py
      -polygon_to_points.py
    +segnet
      +Models
      -pysegnet.py
    -data_collection.py
    -image_recognition.py
    -image_segmentation.py
    -feature_vector_classification.py
  -README.md
  -INSTALL.md

These instructions is about how you copy this project up and running on your local machine for development and testing purposes.

• Python 2.7
• GPU card with CUDA Compute Capability 3.0 or higher NVIDIA's documentation
• Install CUDA Toolkit 8.0
• Download cuDNN v5.1
• Install TensorFlow with GPU
• Install Caffe-Segnet with GPU, caffe Installation ins here

• see Instalation instruction

• see Running instruction

• see Training model and Evaluation

• see Demo

• TensorFlow - Image Recognition
• Caffe-Segnet - Image Segmentation
• XGBoost - Feature Vector Classifier
• Overpass API - Road/Street Geojson
• Google Map API - Street View Image and Visualization