Lightweight and Robust Representation of Economic Scales from Satellite Imagery
Authors: Sungwon Han, Donghyun Ahn, Hyunji Cha, Jeasurk Yang, Sungwon Park, Meeyoung Cha428-436
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | An extensive evaluation demonstrates that the model outperforms state-of-the-art models in predicting economic scales, such as the population density in South Korea (R2=0.9617), and shows a high use potential in developing countries where district-level economic scales are unknown. Results Performance evaluation and ablation study We conduct a set of experiments. The first evaluation takes advantage of the population demographics by dividing these data into a training set and a test set in an 80% 20% ratio. 4fold cross-validation is applied to the training data set to tune the model s hyperparameters, such as the PCA dimensions and the regularization term in the cost function. |
| Researcher Affiliation | Academia | Sungwon Han, 1,3 Donghyun Ahn, 1,3 Hyunji Cha, 1,3 Jeasurk Yang,2,3 Sungwon Park,1,3 Meeyoung Cha3,1 1School of Computing, KAIST, South Korea 2The Institute for Korean Regional Studies, SNU, South Korea 3Data Science Group, IBS, South Korea |
| Pseudocode | No | The paper describes the model's steps and architecture but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | The code is released at Git Hub.1 1https://github.com/Sungwon-Han/READ |
| Open Datasets | Yes | This study utilizes the following data: regional-level demographics and high-resolution satellite imagery. Both data types from many developed countries are accessible via the REST APIs of Esri R Arc GIS, a famous repository of maps and geographic (Johnston et al. 2001). ... The World Imagery satellite data captured by Digital Globe provides 256x256-pixel image tiles... The Esri Demographics by Michael Bauer Research GmbH provides 2018 demographics data and the boundary polygon shapes of districts in 135 countries. |
| Dataset Splits | Yes | The first evaluation takes advantage of the population demographics by dividing these data into a training set and a test set in an 80% 20% ratio. 4fold cross-validation is applied to the training data set to tune the model s hyperparameters, such as the PCA dimensions and the regularization term in the cost function. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU or CPU models, or detailed computer specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions software like XGBoost, ResNet18, DenseNet121, AlexNet, and VGG16, but it does not specify any version numbers for these or other software dependencies. |
| Experiment Setup | Yes | The first evaluation takes advantage of the population demographics by dividing these data into a training set and a test set in an 80% 20% ratio. 4fold cross-validation is applied to the training data set to tune the model s hyperparameters, such as the PCA dimensions and the regularization term in the cost function. and we increase the weight of the unlabeled loss from 0 to 12.5 during the first 40 epochs. and Stochastic gradient descent was used to reduce the loss term, and data-augmentation methods, such as rotating or flipping figures, were used to increase the amount of training data. |