Datasets:
Tasks:
Feature Extraction
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
remote-sensing
aerial-imagery
orthomosaic
lighting-invariance
semantic-stability
vision-encoder
License:
File size: 3,169 Bytes
6441b45 84e4785 059980a 6ebb9c9 84e4785 059980a e3b81b6 6441b45 e3b81b6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
---
dataset_info:
features:
- name: image_t0
dtype: image
- name: image_t1
dtype: image
- name: image_t2
dtype: image
- name: idx
dtype: string
splits:
- name: train
num_bytes: 3569866407.0
num_examples: 609
download_size: 3570065377
dataset_size: 3569866407.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: cc-by-4.0
task_categories:
- feature-extraction
tags:
- remote-sensing
- aerial-imagery
- orthomosaic
- lighting-invariance
- semantic-stability
- vision-encoder
- time-series
size_categories:
- 1K<n<10K
language:
- en
pretty_name: Light Stable Semantics
---
# Light Stable Semantics Dataset
## Dataset Description
This dataset contains aerial orthomosaic tiles captured at three different times of day (10:00, 12:00, and 15:00) to develop vision encoders that are semantically stable under varying lighting conditions. The dataset is designed for training computer vision models that can maintain consistent feature representations despite changes in illumination.
### Dataset Summary
- **Purpose**: Training light-stable semantic vision encoders
- **Data Type**: Aerial orthomosaic tiles (RGBA, 1024x1024 pixels)
- **Time Points**: 3 captures per location (morning, noon, afternoon)
- **Coverage**: Lower Partridge area aerial survey
- **Date**: November 7, 2024 (241107)
- **Location**: MPG Ranch, Montana, USA
### Data Structure
Each record contains:
- `image_t0`: Morning image (10:00, time=1000)
- `image_t1`: Noon image (12:00, time=1200)
- `image_t2`: Afternoon image (15:00, time=1500)
- `idx`: Tile identifier in format `{ROW}_{COL}`
### Data Collection
The orthomosaics were captured using drone surveys with identical geographic bounds but at different times of day to capture varying lighting conditions. All tiles:
- Are 1024x1024 pixels of 1.2cm resolution
- Maintain spatial alignment across time points
- Use consistent geographic coordinates
### Use Cases
This dataset is intended for:
- Training vision encoders robust to lighting changes
- Semantic stability research in computer vision
- Time-invariant feature learning
- Remote sensing applications requiring lighting robustness
## Citation
If you use this dataset in your research, please cite:
```bibtex
@dataset{mpg_ranch_light_stable_semantics_2024,
title={Light Stable Semantics Dataset},
author={Kyle Doherty and Erik Samose and Max Gurinas and Brandon Trabucco and Ruslan Salakhutdinov},
year={2024},
month={November},
url={https://huggingface.co/datasets/mpg-ranch/light-stable-semantics},
publisher={Hugging Face},
note={Aerial orthomosaic tiles captured at multiple times of day for light-stable semantic vision encoder training},
location={MPG Ranch, Montana, USA},
survey_date={2024-11-07},
organization={MPG Ranch}
}
```
## Licensing
This dataset is released under the [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/) license.
Under the following terms:
- **Attribution** — You must give appropriate credit to MPG Ranch, provide a link to the license, and indicate if changes were made. |