development-measure
The official implementation of 'A human-machine collaborative approach measures economic development using satellite imagery'.
Science Score: 26.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
○DOI references
-
○Academic publication links
-
○Committers with academic emails
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (12.4%) to scientific vocabulary
Keywords
Repository
The official implementation of 'A human-machine collaborative approach measures economic development using satellite imagery'.
Basic Info
Statistics
- Stars: 17
- Watchers: 1
- Forks: 2
- Open Issues: 0
- Releases: 3
Topics
Metadata Files
README.md
A human-machine collaborative approach measures economic development using satellite imagery
This repository is official implementation of 'A human-machine collaborative approach measures economic development using satellite imagery'.
Model description
To train the scoring model siScore, you need to follow three stages: Stage1, Stage2, Stage3.
Stage1 first conducts pre-training, and then generates the clusters from given satellite imagery via DeepCluster.
Stage2 suggests ensemble method for aggregating human-guided weak-supervision in reasonable way. Humman annotators first label the partial orders between the clusters from Stage1. Then, the Machine ensembles the all labels from the human, and prune some clusters for better training.
Stage3 trains the rank-wise score model siScore from the Stage2's label (Ensemble POG).
Please refer to README.md in each stage's directory (i.e., Stage1, Stage2, Stage3) to get started.
Dataset in data directory is sampled example from the original dataset. The code is implemented according to its format, so use them just for the reference.
siScore prediction on North Korea
Interactive siScore map example : District-level siScore map on North Korea
Check out for more examples at siscore.app
Original data here (2016~2019)
Visualization of economic development levels predicted by our human-machine collaboration model.
(A) Prediction scores over grid images averaged over four years from 2016 to 2019, (B) shows the yearly aggregated VIIRS nightlight data in 2019, and (C) shows the land cover classification map released by the South Korean Government in 2019. The zoomed-in views in (DF) compare predictions for Sepho County in the Kangwon region. From left to right are the Sentinel-2 satellite images taken in 2019 (D), model predictions (E), and manually verified buildings colored red from the building footprint data in 2014 (F).
Reproducing Figs/Tables
To reproduce the Figures and Tables in the ...
Main text
Supplemetary Materials
please follow the instructions provided at the links above.
Software
Module dependencies
This code has been tested and confirmed to be reproducible with 
Also, all code has been tested on the following environment (as we wrote in requirements.txt):
torch==1.11.0
torchvision==0.8.2
numpy==1.19.2
scipy==1.6.2
sklearn==0.23.2
scikit-image==0.19.2
pandas==1.1.4
geopandas==0.10.2
Pillow==8.2.0
opencv-python==4.7.0.72
faiss-cpu==1.7.3
faiss-gpu==1.7.2
However, certain module versions may cause dependency issues (as of April 11, 2023).
Therefore, I suggest following the command below:
conda install pytorch==1.13.1 torchvision==0.14.1 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install scipy scikit-learn scikit-image geopandas opencv-python faiss-cpu faiss-gpu
The overall progress will take less than 30 minutes for the normal desktop computers.
Hardware
GPU server specification
GPU spec : 4x NVIDIA TITAN Xp
OS : linux 4.4.0-190-generic
Contributor : DA, EL, JY
Owner
- Name: Donghyun Ahn
- Login: DonghyunAhn
- Kind: user
- Location: Daejeon, Korea
- Company: School of Computing, KAIST
- Repositories: 2
- Profile: https://github.com/DonghyunAhn
PhD student @ School of Computing, KAIST
GitHub Events
Total
- Watch event: 4
Last Year
- Watch event: 4
Committers
Last synced: over 2 years ago
Top Committers
| Name | Commits | |
|---|---|---|
| DonghyunAhn | s****a@k****r | 99 |
| eunji | m****1@k****r | 36 |
| mk35471 | 4****1 | 14 |
Committer Domains (Top 20 + Academic)
Issues and Pull Requests
Last synced: over 2 years ago
All Time
- Total issues: 0
- Total pull requests: 1
- Average time to close issues: N/A
- Average time to close pull requests: 9 minutes
- Total issue authors: 0
- Total pull request authors: 1
- Average comments per issue: 0
- Average comments per pull request: 0.0
- Merged pull requests: 1
- Bot issues: 0
- Bot pull requests: 0
Past Year
- Issues: 0
- Pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Issue authors: 0
- Pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0
Top Authors
Issue Authors
Pull Request Authors
- mk35471 (1)