GRAM Dataset

May 26, 2026 · View on GitHub

Labels, pseudo masks, and evaluation results for GRAM

This page provides the dataset used for GRAM (Generalized Region-Aware Mixture-of-Experts), presented at AAAI 2026. For details on the model and methodology, please refer to our paper.

⚠️ Note: Satellite imagery is not included due to licensing restrictions. Images can be accessed via ESRI World Imagery Wayback at zoom level 16 (~1.2 m/px). We used y_x as an image ID.


Source Dataset

Pseudo-label Dataset

Link: Google Drive

Pseudo-labels used for MoE source model training, generated by a semi-supervised segmentation model (For the code and dataset for the SSL segmentation model, refer to this paper). The number of pseudo-labels for each city is as follows. For the detail of dataset, please refer to the appendix of our paper.

ContinentCityTotal Grids
AfricaCairo (Egypt)338,503
AfricaCape Town (South Africa)458,799
AfricaNairobi (Kenya)122,764
AfricaOuagadougou (Burkina Faso)147,546
AsiaColombo (Sri Lanka)96,339
AsiaKarachi (Pakistan)321,073
AsiaMumbai (India)169,407
South AmericaCaracas (Venezuela)97,232
South AmericaMedellín (Colombia)83,256
South AmericaRio de Janeiro (Brazil)353,480
Central AmericaPort-au-Prince (Haiti)29,286
Central AmericaTegucigalpa (Honduras)110,988
Total2,328,673

Target Dataset

Ground-Truth Labels

Manually annotated GT labels for the 3 target cities.

Link: Source

CityLabeled GridsTotal Grids
Dar es Salaam (Tanzania)6,050237,974
Kampala (Uganda)2,62348,696
Maputo (Mozambique)8,863242,963
Total17,536529,633

Results

Evaluation results and performance (mIoU / F1-score) for target cities. Note that the spatial resolution of the released predictions has been reduced due to privacy considerations.

Link: Google Drive

MethodDar es SalaamKampalaMaputo
Vanilla Source0.681 / 0.7920.716 / 0.8140.800 / 0.888
MoE Source0.806 / 0.8850.800 / 0.8810.900 / 0.947
SHOT0.712 / 0.8130.713 / 0.8100.813 / 0.895
TENT0.691 / 0.8000.716 / 0.8140.802 / 0.889
CoTTA0.762 / 0.8530.821 / 0.9000.821 / 0.900
SAR0.700 / 0.8070.748 / 0.8430.804 / 0.890
BeCoTTA0.741 / 0.8360.844 / 0.9110.904 / 0.949
GRAM (Ours)0.859 / 0.9210.870 / 0.9270.907 / 0.951

Citation

@inproceedings{lee2026gram,
  title     = {Generalizable Slum Detection from Satellite Imagery with Mixture-of-Experts},
  author    = {Lee, Sumin and Park, Sungwon and Yang, Jeasurk and Kim, Jihee and Cha, Meeyoung},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year      = {2026}
}