#IA Track - Technology and Tools
The volume of Earth Observation imagery, acquired by satellites and drones is huge (~100 To/day).
DeepLearning approach allow to switch from pixels to insights, the only way to be able to really use these initial raw data.
RoboSat.pink is a modular and extensible OpenSource Computer Vision framework, bridging cuting-edges research papers to a robust implementation.
This presentation will focus on:
- Computer Vision 101 reminder
- The specificities of geospatial data aka "Spatial is Special"
- Mains RoboSat.pink use cases
- Current research issues in Computer Vision GeoSpatial
5. THE PERCEPTRON
A PROBABILISTIC MODEL FOR INFORMATION STORAGE
AND ORGANIZATION IN THE BRAIN
Rosenblatt 1958
AI is an old Lady
https://en.wikipedia.org/wiki/Perceptron
37. So all you need is :
- Imagery → any file format readable by GDAL
- GPU → NVIDIA > 8Go RAM
- Initial skills → GeoSpatial Data and CLI fluency
- Labels → usualy the key point
43. How to scale it, or improve it again ?
rsp train add more GPU,
reduce dataset redundancy,
improve model, loss or optimizer
rsp tile add more CPU
use raster compression
rsp predict export model to ONNX or JIT,
then use an high performance inference solution.
47. Request For Funding
- Increase prediction accuracy :
- on low resolution imagery
- even with few labels
- on network features
- instance segmentation
48. Request For Funding
- Increase prediction accuracy :
- on low resolution imagery
- even with few labels
- on network features
- instance segmentation
- Improve performances
- on Training
- on Inference
49. Request For Funding
- Increase prediction accuracy :
- on low resolution imagery
- even with few labels
- on network features
- instance segmentation
- Improve performances
- on Training
- on Inference
- Add support for :
- MultiClass classification
- Time Series Imagery
50. Why using Deep Learning for GeoSpatial ?
Easy to spotify at scale inconsistencies beetwen two datasets
51. Why using Deep Learning for GeoSpatial ?
Easy to spotify at scale inconsistencies beetwen two datasets
If you provide accurate labels matching an imagery,
infere at scale on similar new imageries
52. Why using Deep Learning for GeoSpatial ?
Easy to spotify at scale inconsistencies beetwen two datasets
If you provide accurate labels matching an imagery,
infere at scale on similar new imageries
But prefer Machine Learning if your classification is trivial
54. Why using RoboSat.pink ?GIS Standards compliancy
Ease Data Preparation
Build-in WebUI
Handle MultiBands Imagery and DataFusion
High Performances
Easy to deploy
Accurate (state of art Computer Vision)
Extensible by design
Open Source
Community
59. Extract insights from GeoSpatial data with Deep Learning
@data_pink
www.datapink.com
60. RoboSat.pink powered by @data_pink
Take Away
- Industrial AI4EO Imagery framework available
- Performances already OK to use it for region or country
- No need anymore to be a Computer Vision expert to use it
- Plain OpenData can be use to train accurate model
- Funding and Pull Requests can make the difference