When we train models on biased datasets, they not only reproduce data biases, but can worsen them at test time — a phenomenon called bias amplification. Many of the current bias amplification metrics (e.g., BA→, DPA) measure bias amplification only in classification datasets. These metrics are ineffective for image captioning datasets, as they cannot capture the language semantics of a caption. Recent work introduced Leakage in Captioning (LIC), a language-aware bias amplification metric that understands caption semantics. However, LIC has a crucial limitation: it cannot identify the source of bias amplification in captioning models. We propose Directional Bias Amplification in Captioning (DBAC), a language-aware and directional metric that can identify when captioning models amplify biases. DBAC has two more improvements over LIC: (1) it is less sensitive to sentence encoders (a hyperparameter in language-aware metrics), and (2) it provides a more accurate estimate of bias amplification in captions. Our experiments on gender and race attributes in the COCO captions dataset show that DBAC is the only reliable metric to measure bias amplification in captions.

Deep learning models commonly benefit from data augmentation techniques to diversify the set of training images. When working with satellite imagery, it is common for practitioners to apply a limited set of transformations developed for natural images (e.g., flip and rotate) to expand the training set without overly modifying the satellite images. There are many techniques for natural image data augmentation, but given the differences between the two domains, it is not clear whether data augmentation methods developed for natural images are well suited for satellite imagery. This paper presents an extensive experimental study on three classification and three regression tasks over four satellite image datasets. We compare common computer vision data augmentation techniques and propose three novel satellite-specific data augmentation strategies. Across tasks and datasets, we find that geometric transformations are beneficial for satellite imagery while color transformations generally are not. Additionally, our novel Sat-SlideMix, Sat-CutMix, and Sat-Trivial methods all exhibit strong performance across all tasks and datasets.

Crop field boundaries are foundational datasets for agricultural monitoring and assessments but are expensive to collect manually. Machine learning (ML) methods for automatically extracting field boundaries from remotely sensed images could help realize the demand for these datasets at a global scale. However, current ML methods for field instance segmentation lack sufficient geographic coverage, accuracy, and generalization capabilities. Further, research on improving ML methods is restricted by the lack of labeled datasets representing the diversity of global agricultural fields. We present Fields of The World (FTW)---a novel ML benchmark dataset for agricultural field instance segmentation spanning 24 countries on four continents (Europe, Africa, Asia, and South America). FTW is an order of magnitude larger than previous datasets with 70,462 samples, each containing instance and semantic segmentation masks paired with multi-date, multi-spectral Sentinel-2 satellite images. We provide results from baseline models for the new FTW benchmark, show that models trained on FTW have better zero-shot and fine-tuning performance in held-out countries than models that aren't pre-trained with diverse datasets, and show positive qualitative zero-shot results of FTW models in a real-world scenario -- running on Sentinel-2 scenes over Ethiopia.

Geologists seek to understand the relationship between volcanic unrest and eruptions by identifying subtle Volcanic Thermal Features (VTFs) in high-resolution satellite imagery. This analysis requires the careful curation of large databases of relevant volcanic thermal information. However, volcanic unrest is characterized by highly subtle thermal anomalies. Manual identification on a global scale is highly labor- and time-intensive. We propose Hotspotter: an end-to-end system to automatically detect subtle volcanic thermal anomalies in satellite images and derive relevant thermal statistics. Previous solutions for automated VTF detection have limited data size and geographic diversity. To accommodate an unprecedentedly large and diverse volcanic dataset, we propose an automated pipeline combining unsupervised anomaly detection with supervised classification to filter anomalous regions. Hotspotter gives 90% anomaly detection accuracy and robust generalization to new volcanoes. Our automated approach can accelerate scientists' search for VTFs to help identify relevant thermal precursors and enable more precise forecasts of global volcanic eruptions.

We introduce a highly multimodal transformer to represent many remote sensing modalities - multispectral optical, synthetic aperture radar, elevation, weather, pseudo-labels, and more - across space and time. These inputs are useful for diverse remote sensing tasks, such as crop mapping and flood detection. However, learning shared representations of remote sensing data is challenging, given the diversity of relevant data modalities, and because objects of interest vary massively in scale, from small boats (1-2 pixels and fast) to glaciers (thousands of pixels and slow). We present a novel self-supervised learning algorithm that extracts multi-scale features across a flexible set of input modalities through masked modeling. Our dual global and local contrastive losses differ in their targets (deep representations vs. shallow input projections) and masking strategies (structured vs. not). Our Galileo is a single generalist model that outperforms SoTA specialist models for satellite images and pixel time series across eleven benchmarks and multiple tasks.

Foundation models have enabled rapid progress across many specialized domains by leveraging large-scale pre-training on unlabeled data, demonstrating strong generalization to a variety of downstream tasks. While such models have gained significant attention in fields like Earth Observation, their application to Mars science remains limited. A key enabler of progress in other domains has been the availability of standardized benchmarks that support systematic evaluation. In contrast, Mars science lacks such benchmarks and standardized evaluation frameworks, which have limited progress toward developing foundation models for Martian tasks. To address this gap, we introduce Mars-Bench, the first benchmark designed to systematically evaluate models across a broad range of Mars-related tasks using both orbital and surface imagery. Mars-Bench comprises 20 datasets spanning classification, segmentation, and object detection, focused on key geologic features such as craters, cones, boulders, and frost. We provide standardized, ready-to-use datasets and baseline evaluations using models pre-trained on natural images, Earth satellite data, and state-of-the-art vision-language models. Results from all analyses suggest that Mars-specific foundation models may offer advantages over general-domain counterparts, motivating further exploration of domain-adapted pre-training. Mars-Bench aims to establish a standardized foundation for developing and comparing machine learning models for Mars science. Our data, models, and code are available at: https://mars-bench.github.io/.

Most ML datasets today contain biases. When we train models on these datasets, they often not only learn these biases but can worsen them --- a phenomenon known as bias amplification. Several co-occurrence-based metrics have been proposed to measure bias amplification in classification datasets. They measure bias amplification between a protected attribute (e.g., gender) and a task (e.g., cooking). These metrics also support fine-grained bias analysis by identifying the direction in which a model amplifies biases. However, co-occurrence-based metrics have limitations --- some fail to measure bias amplification in balanced datasets, while others fail to measure negative bias amplification. To solve these issues, recent work proposed a predictability-based metric called leakage amplification (LA). However, LA cannot identify the direction in which a model amplifies biases. We propose Directional Predictability Amplification (DPA), a predictability-based metric that is (1) directional, (2) works with balanced and unbalanced datasets, and (3) correctly identifies positive and negative bias amplification. DPA eliminates the need to evaluate models on multiple metrics to verify these three aspects. DPA also improves over prior predictability-based metrics like LA: it is less sensitive to the choice of attacker function (a hyperparameter in predictability-based metrics), reports scores within a bounded range, and accounts for dataset bias by measuring relative changes in predictability. Our experiments on well-known datasets like COMPAS (a tabular dataset), COCO, and ImSitu (image datasets) show that DPA is the most reliable metric to measure bias amplification in classification problems.

Satellite data has the potential to inspire a seismic shift for machine learning—one in which we rethink existing practices designed for traditional data modalities. As machine learning for satellite data (SatML) gains traction for its real-world impact, our field is at a crossroads. We can either continue applying ill-suited approaches, or we can initiate a new research agenda that centers around the unique characteristics and challenges of satellite data. This position paper argues that satellite data constitutes a distinct modality for machine learning research and that we must recognize it as such to advance the quality and impact of SatML research across theory, methods, and deployment. We outline research directions, critical discussion questions and actionable suggestions to transform SatML from merely an intriguing application area to a dedicated research discipline that helps move the needle on big challenges for machine learning and society.

In this position paper, we argue that application-driven research has been systemically under-valued in the machine learning community. As applications of machine learning proliferate, innovative algorithms inspired by specific real-world challenges have become increasingly important. Such work offers the potential for significant impact not merely in domains of application but also in machine learning itself. In this paper, we describe the paradigm of application-driven research in machine learning, contrasting it with the more standard paradigm of methods-driven research. We illustrate the benefits of application-driven machine learning and how this approach can productively synergize with methods-driven work. Despite these benefits, we find that reviewing, hiring, and teaching practices in machine learning often hold back application-driven innovation. We outline how these processes may be improved.

Over the years, space scientists have collected terabytes of Mars data from satellites and rovers. One important set of features identified in Mars orbital images is pitted cones, which are interpreted to be mud volcanoes believed to form in regions that were once saturated in water (i.e., a lake or ocean). Identifying pitted cones globally on Mars would be of great importance, but expert geologists are unable to sort through the massive orbital image archives to identify all examples. However, this task is well suited for computer vision. Although several computer vision datasets exist for various Mars-related tasks, there is currently no open-source dataset available for cone detection/segmentation. Furthermore, previous studies trained models using data from a single region, which limits their applicability for global detection and mapping. Motivated by this, we introduce ConeQuest, the first expert-annotated public dataset to identify cones on Mars. ConeQuest consists of >13k samples from 3 different regions of Mars. We propose two benchmark tasks using ConeQuest: (i) Spatial Generalization and (ii) Cone-size Generalization. We finetune and evaluate widely-used segmentation models on both benchmark tasks. Results indicate that cone segmentation is a challenging open problem not solved by existing segmentation models, which achieve an average IoU of 52.52% and 42.55% on in-distribution data for tasks (i) and (ii), respectively. We believe this new benchmark dataset will facilitate the development of more accurate and robust models for cone segmentation. Data and code are available at https://github.com/kerner-lab/ConeQuest.

The desired output for most real-world tasks using machine learning (ML) and remote sensing data is a set of dense predictions that form a predicted map for a geographic region. However, most prior work involving ML and remote sensing follows the traditional practice of reporting metrics on a set of independent, geographically-sparse samples and does not perform dense predictions. To reduce the labor of producing dense prediction maps, we present OpenMapFlow---an open-source python library for rapid map creation with ML and remote sensing data. OpenMapFlow provides 1) a data processing pipeline for users to create labeled datasets for any region, 2) code to train state-of-the-art deep learning models on custom or existing datasets, and 3) a cloud-based architecture to deploy models for efficient map prediction. We demonstrate the benefits of OpenMapFlow through experiments on three binary classification tasks: cropland, crop type (maize), and building mapping. We show that OpenMapFlow drastically reduces the time required for dense prediction compared to traditional workflows. We hope this library will stimulate novel research in areas such as domain shift, unsupervised learning, and societally-relevant applications and lessen the barrier to adopting research methods for real-world tasks.

Recent progress in self-supervision has shown that pre-training large neural networks on vast amounts of unsupervised data can lead to substantial increases in generalization to downstream tasks. Such models, recently coined foundation models, have been transformational to the field of natural language processing.Variants have also been proposed for image data, but their applicability to remote sensing tasks is limited.To stimulate the development of foundation models for Earth monitoring, we propose a benchmark comprised of six classification and six segmentation tasks, which were carefully curated and adapted to be both relevant to the field and well-suited for model evaluation. We accompany this benchmark with a robust methodology for evaluating models and reporting aggregated results to enable a reliable assessment of progress. Finally, we report results for 20 baselines to gain information about the performance of existing models.We believe that this benchmark will be a driver of progress across a variety of Earth monitoring tasks.

Analyzing remote sensing datasets to identify sites of interest for in-situ investigation in robotic and human exploration is challenging with current methods. In this study, we investigated the utility of machine learning algorithms for outlier detection to identify samples of interest at field exploration sites based on remotely-sensed observations. In our analysis of two study sites, Cucomungo alluvial fan in Death Valley, CA and Jezero crater on Mars, we found that outliers identified by the isolation forest algorithm in satellite datasets seemed to correspond to unique compositions or surface properties. Our results show that outlier detection is a promising method for assisting scientists by guiding field exploration site selection.

Measures to slow the spread of COVID-19 are affecting the food supply chain in many ways including the availability of inputs, labor, transport, and cross-border trade. Earth Observation (EO) from satellites can provide timely and transparent evidence on the extent and impact these measures have on the agricultural activities and related food systems. EO capabilities required to address information needs related to global food supply, national harvesting statistics, national relief programs and labor-intensive crop production were made available over the trilateral COVID-19 Earth Observing Dashboard. The presented use cases make full use of the combined satellites fleet of NASA, ESA and JAXA as well as the expertise of the EO community.

Abstract Science teams for rover-based planetary exploration missions like the Mars Science Laboratory Curiosity rover have limited time for analyzing new data before making decisions about follow-up observations. There is a need for systems that can rapidly and intelligently extract information from planetary instrument datasets and focus attention on the most promising or novel observations. Several novelty detection methods have been explored in prior work for three-channel color images and non-image datasets, but few have considered multispectral or hyperspectral image datasets for the purpose of scientific discovery. We compared the performance of four novelty detection methods—Reed Xiaoli (RX) detectors, principal component analysis (PCA), autoencoders, and generative adversarial networks (GANs)—and the ability of each method to provide explanatory visualizations to help scientists understand and trust predictions made by the system. We show that pixel-wise RX and autoencoders trained with structural similarity (SSIM) loss can detect morphological novelties that are not detected by PCA, GANs, and mean squared error autoencoders, but that the latter methods are better suited for detecting spectral novelties—i.e., the best method for a given setting depends on the type of novelties that are sought. Additionally, we find that autoencoders provide the most useful explanatory visualizations for enabling users to understand and trust model detections, and that existing GAN approaches to novelty detection may be limited in this respect.

In this work, we present a system based on convolutional autoencoders for detecting novel features in multispectral images. We introduce SAMMIE: Selections based on Autoencoder Modeling of Multispectral Image Expectations. Previous work using autoencoders employed the scalar reconstruction error to classify new images as novel or typical. We show that a spatial-spectral error map can enable both accurate classification of novelty in multispectral images as well as human-comprehensible explanations of the detection. We apply our methodology to the detection of novel geologic features in multispectral images of the Martian surface collected by the Mastcam imaging system on the Mars Science Laboratory Curiosity rover.