Skip to content

Latest commit

 

History

History
113 lines (87 loc) · 3.66 KB

README.md

File metadata and controls

113 lines (87 loc) · 3.66 KB
Raw

Description

This is going to describe all of processes, start from very begining to the end of process.

1. Data Collection

Step 0: Convert shape file to geojson file

Command line to convert .shp to .geojson link to install

  • input: subdistrict.shp
  • output: subdistrict.geojson

For example,

ogr2ogr -f GeoJSON output.geojson input.shp

Step 1: Convert polygon's subdistrict to points/coordinate

image_retreival/polygon_to_points.py

  • input: subdistrict.geojson, province, district, and subdist
  • output: points/coordinate of subdistrict

Step 2: Extract village points from subdistrict points

image_retreival/get_village_points.py

  • input: points of subdistrict
  • output: points of village

Step 3: Load Google Street View Image from village points

image_retreival/GSV_loader.py

  • input: points of village
  • output: street view images

2. Data Processing

Step 1: Classify image

  • input: street view images
  • output: features vector (.csv)

Step 1a: Image Segmentation

image_segmentation.py

  • input: google street view images
  • output: segmented image

Step 1b: Image Recognition

image_recognition.py

  • input: street view and segmented images
  • output: top-5 classification results (.csv)

Combined Segmentation and Recognition

Step 2: Features Vector Classification

  • input: features (.csv)
  • output: classification results (.csv)

** Example of classification results **

Image name 1st result 1st score 2nd result 2nd score 3rd result 3rd score 4th result 4th score 5th result 5th score
8.076112623778412-98.83164102093671-240-2016-03-12-Pavement pot, flowerpot 0.22256 sundial 0.0321534 mortar 0.0310492 caldron, cauldron 0.0249833 Dutch oven 0.0229676
8.078046220141706-98.85250783470164-240_2016-03-8-Fence greenhouse 0.28458 steel arch bridge 0.0468726 worm fence 0.0316026 valley, vale 0.0296596 seashore, coast 0.015028

3. Data Visualization

Step 1: Generate geojson from classification result

feature_vector/feature_vector_classification.py

  • input: classificatino results (.csv)
  • output: geojson (.js)

Step 2: Visualize on map

GSV/../../../visulization.py

  • input: geojson (.js)
  • output: map visualization

How to run

Assuming you have installed all of require libraries in Installing, simply run

  • Data Collection

python data_collection.py --province PROVINCE --district DISTRICT --subdist SUBDIST --village VILLAGE

  • Data Processing

python image_recognition.py --province PROVINCE --district DISTRICT --subdist SUBDIST --village VILLAGE
python image_segmentation.py --province PROVINCE --district DISTRICT --subdist SUBDIST --village VILLAGE
python feature_vector_classification.py --province PROVINCE --district DISTRICT --subdist SUBDIST --village VILLAGE