THE PLATFORM

the DYMAX platform

Using our fully cloud-enabled platform, private companies and the public sector accelerate strategic data-driven decisions from their geospatial imagery targets.

how does it work?

01. problem

Problem definition

Available data

Alternate sources

02. data

Data wrangling

Preprocessing

Outlier removal

03. annotation

Imagery labeling

Experts validation

Ground truth

04. modeling

Train/ Test split

Deep Neural Network

Hyperparameters

05. deploy

Map tile server

Data compression

Visualization

use cases

agro › crop monitoring

We developed with BAYER a model to monitor how the crop is responding to normal and extreme flood and drought situations. Our DYMAX platform combined data from SAT/SAR imagery, NASA flood historical data and BAYER’s proprietary information.

 

This model brought more granularity to the data, so BAYER can discover new patterns by state, county and lot.

 

Our team provided execution, custom development, testing, and model validation. We improved levels of accuracy having kappa_index>0.95

 

Outcome I: Generation of monthly analytics about harvest performance, by crop, by lot.

Outcome II: Discover and forecast new lots for commercial development

urban › illegal mining detection

Environmental impacts of mining can occur at local, regional, and global scales through direct and indirect mining practices. Impacts can result in erosion, sinkholes, loss of biodiversity, or the contamination of soil, groundwater, and surface water by the chemicals emitted from mining processes. Dymaxion Labs teamed up with PNUD and EbA Lomas to create a model to detect illegal mining patterns, having the city of Lima (Peru) as the initial target.

 

Our DYMAX platform combined optical data from Sentinel 2, with layers of elevation maps and weather historical data to train a custom model. Due to cloudy conditions, we also enabled the SAR acquisition process to include imagery from Sentinel 1.

 

Outcome I: Discover new areas of illegal mining.

Outcome II: Integration with IOT sensors to gather real-time weather variables.

Outcome III: Develop an early warning detection system, triggering alarms.

Outcome IV: Time series forecasting using recurrent neural networks (RNN)

 

Proyecto EbA Lomas: 

https://www.pe.undp.org/content/peru/es/home/operations/projects/environment_and_energy/eba-lomas.html

climate › deforestation patterns

Land invasions an fast urban growth generate a huge deforestation process. We partied with IDB Bank and Manaus sub-Secretariat of Information Technology to gather drone imagery to understand deforestation patterns originated by development of informal settlements. The city of Manaus is located in the center of Amazonas, Brazil, being the world’s largest rainforest. Key figures: population 2.2M, area of 11,000 km2.

 

The DYMAX platform, combining drone imagery, AI, and ML algorithms served Manaus, to integrate informal dwellers into social programs of employment, income, ando housing, and promote land upgrading, infrastructure mapping, tax planning, enviromental monitoring, ando others, which are essential features for sustainable urban development.

 

 

Press release: https://blogs.iadb.org/ciudades-sostenibles/en/monitoring-informal-settlement-growth-in-manaus-brazil-with-drones/