Search Results for author: Freddie Kalaitzis

Found 13 papers, 5 papers with code

Exploring DINO: Emergent Properties and Limitations for Synthetic Aperture Radar Imagery

no code implementations5 Oct 2023 Joseph A. Gallego-Mejia, Anna Jungbluth, Laura Martínez-Ferrer, Matt Allen, Francisco Dorr, Freddie Kalaitzis, Raúl Ramos-Pollán

We observe a small improvement in model performance with pre-training compared to training from scratch and discuss the limitations and opportunities of SSL for remote sensing and land cover segmentation.

Image Segmentation Segmentation +2

Exploring Generalisability of Self-Distillation with No Labels for SAR-Based Vegetation Prediction

no code implementations3 Oct 2023 Laura Martínez-Ferrer, Anna Jungbluth, Joseph A. Gallego-Mejia, Matt Allen, Francisco Dorr, Freddie Kalaitzis, Raúl Ramos-Pollán

In this work we pre-train a DINO-ViT based model using two Synthetic Aperture Radar datasets (S1GRD or GSSIC) across three regions (China, Conus, Europe).

Fewshot learning on global multimodal embeddings for earth observation tasks

no code implementations29 Sep 2023 Matt Allen, Francisco Dorr, Joseph A. Gallego-Mejia, Laura Martínez-Ferrer, Anna Jungbluth, Freddie Kalaitzis, Raúl Ramos-Pollán

In this work we pretrain a CLIP/ViT based model using three different modalities of satellite imagery across five AOIs covering over ~10\% of Earth's total landmass, namely Sentinel 2 RGB optical imagery, Sentinel 1 SAR radar amplitude and interferometric coherence.

Enhancing Self-Supervised Learning for Remote Sensing with Elevation Data: A Case Study with Scarce And High Level Semantic Labels

1 code implementation13 Apr 2023 Omar A. Castaño-Idarraga, Raul Ramos-Pollán, Freddie Kalaitzis

We assess the performance of our approach on a binary semantic segmentation task and a binary image classification task, both derived from a dataset created for the northwest of Colombia.

Binary Classification Image Classification +2

Deep learning based landslide density estimation on SAR data for rapid response

no code implementations18 Nov 2022 Vanessa Boehm, Wei Ji Leong, Ragini Bal Mahesh, Ioannis Prapas, Edoardo Nemni, Freddie Kalaitzis, Siddha Ganju, Raul Ramos-Pollan

Since such data might not be available during other events or regions, we aimed to produce a landslide density map using only elevation and SAR data to be useful to decision-makers in rapid response scenarios.

Density Estimation

SAR-based landslide classification pretraining leads to better segmentation

1 code implementation17 Nov 2022 Vanessa Böhm, Wei Ji Leong, Ragini Bal Mahesh, Ioannis Prapas, Edoardo Nemni, Freddie Kalaitzis, Siddha Ganju, Raul Ramos-Pollan

In the case of landslides, rapid assessment involves determining the extent of the area affected and measuring the size and location of individual landslides.

Classification Landslide segmentation

Deep Learning for Rapid Landslide Detection using Synthetic Aperture Radar (SAR) Datacubes

1 code implementation5 Nov 2022 Vanessa Boehm, Wei Ji Leong, Ragini Bal Mahesh, Ioannis Prapas, Edoardo Nemni, Freddie Kalaitzis, Siddha Ganju, Raul Ramos-Pollan

With climate change predicted to increase the likelihood of landslide events, there is a growing need for rapid landslide detection technologies that help inform emergency responses.

Open High-Resolution Satellite Imagery: The WorldStrat Dataset -- With Application to Super-Resolution

1 code implementation13 Jul 2022 Julien Cornebise, Ivan Oršolić, Freddie Kalaitzis

We hereby hope to foster broad-spectrum applications of ML to satellite imagery, and possibly develop from free public low-resolution Sentinel2 imagery the same power of analysis allowed by costly private high-resolution imagery.

Humanitarian Multi-Frame Super-Resolution

Cannot find the paper you are looking for? You can Submit a new open access paper.