First place solution for Crop Detection from Satellite Imagery competition organized by CV4A workshop at ICLR 2020. The model architecture consists of 3-layer Conv-net, Masked Features Averaging layer, 3-layer Bi-directional GRU-net and fully connected classification layer. Masked Features Averaging layer is similar to global average pooling but only averages pixels belong to crop field. read more...

This is a PyTorch model trained on the Tropical Cyclone Wind Estimation Competition dataset with v0.1 of the TorchGeo package. The model is a resnet18 model pretrained on ImageNet then trained with a MSE loss. The data were randomly split 80/20 by storm ID and an early stop was used based on performance. read more...

This model classifies crop types for each field based on the field as well as on its surroundings. In the Zindi AgriFieldNet India Challenge this was the third place solution by the team re-union in the final round to classify crop types in agricultural fields across Northern India using multispectral observations from Sentinel-2 satellite. read more...

Small farms produce about 35% of the worldʼs food, and are mostly found in low-and middle-income countries. Reliable information about these farms is limited, making support and policy-making difficult. Earth Observation data from satellites such as Sentinel-2, in combination with machine learning, can help improve agricultural monitoring, crop mapping, and disaster risk management for these small farms. The Main goal of this challenge is to classify crop types in agricultural fields across Northern India using multispectral observations from Sentinel-2 satellite. These fields are located in various districts in states of Uttar Pradesh, Rajasthan, Odisha and Bihar. read more...

Second place solution to classify crop types in agricultural fields across Northern India using multispectral observations from Sentinel-2 satellite. Ensembled weighted tree-based models "LGBM, CATBOOST, XGBOOST" with stratified k-fold cross validation, taking advantage of spatial variability around each field within different distances. read more...

The Replicable AI for Microplanning (Ramp) deep learning model is a semantic segmentation one which detects buildings from satellite imagery and delineates the footprints in low-and-middle-income countries (LMICs) using satellite imagery and enables in-country users to build their own deep learning models for their regions of interest. The architecture and approach were inspired by the Eff-UNet model outlined in this CVPR 2020 Paper. read more...