Siddha Ganju

Vanessa Boehm, Wei Ji Leong, Ragini Bal Mahesh et al. · cmu, nvidia

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. Synthetic Aperture Radar (SAR) is a remote sensing technique that can provide measurements of affected areas independent of weather or lighting conditions. Usage of SAR, however, is hindered by domain knowledge that is necessary for the pre-processing steps and its interpretation requires expert knowledge. We provide simplified, pre-processed, machine-learning ready SAR datacubes for four globally located landslide events obtained from several Sentinel-1 satellite passes before and after a landslide triggering event together with segmentation maps of the landslides. From this dataset, using the Hokkaido, Japan datacube, we study the feasibility of SAR-based landslide detection with supervised deep learning (DL). Our results demonstrate that DL models can be used to detect landslides from SAR data, achieving an Area under the Precision-Recall curve exceeding 0.7. We find that additional satellite visits enhance detection performance, but that early detection is possible when SAR data is combined with terrain information from a digital elevation model. This can be especially useful for time-critical emergency interventions. Code is made publicly available at https://github.com/iprapas/landslide-sar-unet.

Vanessa Böhm, Wei Ji Leong, Ragini Bal Mahesh et al. · cmu, nvidia

Rapid assessment after a natural disaster is key for prioritizing emergency resources. In the case of landslides, rapid assessment involves determining the extent of the area affected and measuring the size and location of individual landslides. Synthetic Aperture Radar (SAR) is an active remote sensing technique that is unaffected by weather conditions. Deep Learning algorithms can be applied to SAR data, but training them requires large labeled datasets. In the case of landslides, these datasets are laborious to produce for segmentation, and often they are not available for the specific region in which the event occurred. Here, we study how deep learning algorithms for landslide segmentation on SAR products can benefit from pretraining on a simpler task and from data from different regions. The method we explore consists of two training stages. First, we learn the task of identifying whether a SAR image contains any landslides or not. Then, we learn to segment in a sparsely labeled scenario where half of the data do not contain landslides. We test whether the inclusion of feature embeddings derived from stage-1 helps with landslide detection in stage-2. We find that it leads to minor improvements in the Area Under the Precision-Recall Curve, but also to a significantly lower false positive rate in areas without landslides and an improved estimate of the average number of landslide pixels in a chip. A more accurate pixel count allows to identify the most affected areas with higher confidence. This could be valuable in rapid response scenarios where prioritization of resources at a global scale is important. We make our code publicly available at https://github.com/VMBoehm/SAR-landslide-detection-pretraining.

Vishal Upendran, Panagiotis Tigas, Banafsheh Ferdousi et al. · cmu, nvidia

Geomagnetically Induced Currents (GICs) arise from spatio-temporal changes to Earth's magnetic field which arise from the interaction of the solar wind with Earth's magnetosphere, and drive catastrophic destruction to our technologically dependent society. Hence, computational models to forecast GICs globally with large forecast horizon, high spatial resolution and temporal cadence are of increasing importance to perform prompt necessary mitigation. Since GIC data is proprietary, the time variability of horizontal component of the magnetic field perturbation (dB/dt) is used as a proxy for GICs. In this work, we develop a fast, global dB/dt forecasting model, which forecasts 30 minutes into the future using only solar wind measurements as input. The model summarizes 2 hours of solar wind measurement using a Gated Recurrent Unit, and generates forecasts of coefficients which are folded with a spherical harmonic basis to enable global forecasts. When deployed, our model produces results in under a second, and generates global forecasts for horizontal magnetic perturbation components at 1-minute cadence. We evaluate our model across models in literature for two specific storms of 5 August 2011 and 17 March 2015, while having a self-consistent benchmark model set. Our model outperforms, or has consistent performance with state-of-the-practice high time cadence local and low time cadence global models, while also outperforming/having comparable performance with the benchmark models. Such quick inferences at high temporal cadence and arbitrary spatial resolutions may ultimately enable accurate forewarning of dB/dt for any place on Earth, resulting in precautionary measures to be taken in an informed manner.

Jonathan Francis, Bingqing Chen, Siddha Ganju et al. · cmu, nvidia

We present the results of our autonomous racing virtual challenge, based on the newly-released Learn-to-Race (L2R) simulation framework, which seeks to encourage interdisciplinary research in autonomous driving and to help advance the state of the art on a realistic benchmark. Analogous to racing being used to test cutting-edge vehicles, we envision autonomous racing to serve as a particularly challenging proving ground for autonomous agents as: (i) they need to make sub-second, safety-critical decisions in a complex, fast-changing environment; and (ii) both perception and control must be robust to distribution shifts, novel road features, and unseen obstacles. Thus, the main goal of the challenge is to evaluate the joint safety, performance, and generalisation capabilities of reinforcement learning agents on multi-modal perception, through a two-stage process. In the first stage of the challenge, we evaluate an autonomous agent's ability to drive as fast as possible, while adhering to safety constraints. In the second stage, we additionally require the agent to adapt to an unseen racetrack through safe exploration. In this paper, we describe the new L2R Task 2.0 benchmark, with refined metrics and baseline approaches. We also provide an overview of deployment, evaluation, and rankings for the inaugural instance of the L2R Autonomous Racing Virtual Challenge (supported by Carnegie Mellon University, Arrival Ltd., AICrowd, Amazon Web Services, and Honda Research), which officially used the new L2R Task 2.0 benchmark and received over 20,100 views, 437 active participants, 46 teams, and 733 model submissions -- from 88+ unique institutions, in 58+ different countries. Finally, we release leaderboard results from the challenge and provide description of the two top-ranking approaches in cross-domain model transfer, across multiple sensor configurations and simulated races.

Siddha Ganju, Amartya Hatua, Peter Jenniskens et al. · cmu, nvidia

The Cameras for Allsky Meteor Surveillance (CAMS) project, funded by NASA starting in 2010, aims to map our meteor showers by triangulating meteor trajectories detected in low-light video cameras from multiple locations across 16 countries in both the northern and southern hemispheres. Its mission is to validate, discover, and predict the upcoming returns of meteor showers. Our research aimed to streamline the data processing by implementing an automated cloud-based AI-enabled pipeline and improve the data visualization to improve the rate of discoveries by involving the public in monitoring the meteor detections. This article describes the process of automating the data ingestion, processing, and insight generation using an interpretable Active Learning and AI pipeline. This work also describes the development of an interactive web portal (the NASA Meteor Shower portal) to facilitate the visualization of meteor radiant maps. To date, CAMS has discovered over 200 new meteor showers and has validated dozens of previously reported showers.

Vanessa Boehm, Wei Ji Leong, Ragini Bal Mahesh et al. · cmu, nvidia

This work aims to produce landslide density estimates using Synthetic Aperture Radar (SAR) satellite imageries to prioritise emergency resources for rapid response. We use the United States Geological Survey (USGS) Landslide Inventory data annotated by experts after Hurricane María in Puerto Rico on Sept 20, 2017, and their subsequent susceptibility study which uses extensive additional information such as precipitation, soil moisture, geological terrain features, closeness to waterways and roads, etc. 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. The USGS Landslide Inventory contains the coordinates of 71,431 landslide heads (not their full extent) and was obtained by manual inspection of aerial and satellite imagery. It is estimated that around 45\% of the landslides are smaller than a Sentinel-1 typical pixel which is 10m $\times$ 10m, although many are long and thin, probably leaving traces across several pixels. Our method obtains 0.814 AUC in predicting the correct density estimation class at the chip level (128$\times$128 pixels, at Sentinel-1 resolution) using only elevation data and up to three SAR acquisitions pre- and post-hurricane, thus enabling rapid assessment after a disaster. The USGS Susceptibility Study reports a 0.87 AUC, but it is measured at the landslide level and uses additional information sources (such as proximity to fluvial channels, roads, precipitation, etc.) which might not regularly be available in an rapid response emergency scenario.

Tarun Narayanan, Ajay Krishnan, Anirudh Koul et al. · cmu, nvidia

Applying Machine learning to domains like Earth Sciences is impeded by the lack of labeled data, despite a large corpus of raw data available in such domains. For instance, training a wildfire classifier on satellite imagery requires curating a massive and diverse dataset, which is an expensive and time-consuming process that can span from weeks to months. Searching for relevant examples in over 40 petabytes of unlabelled data requires researchers to manually hunt for such images, much like finding a needle in a haystack. We present a no-code end-to-end pipeline, Curator, which dramatically minimizes the time taken to curate an exhaustive labeled dataset. Curator is able to search massive amounts of unlabelled data by combining self-supervision, scalable nearest neighbor search, and active learning to learn and differentiate image representations. The pipeline can also be readily applied to solve problems across different domains. Overall, the pipeline makes it practical for researchers to go from just one reference image to a comprehensive dataset in a diminutive span of time.

Brad Larson, Bishal Upadhyaya, Luke McDermott et al. · cmu, nvidia

Deep learning harnesses massive parallel floating-point processing to train and evaluate large neural networks. Trends indicate that deeper and larger neural networks with an increasing number of parameters achieve higher accuracy than smaller neural networks. This performance improvement, which often requires heavy compute for both training and evaluation, eventually needs to translate well to resource-constrained hardware for practical value. Structured pruning asserts that while large networks enable us to find solutions to complex computer vision problems, a smaller, computationally efficient sub-network can be derived from the large neural network that retains model accuracy but significantly improves computational efficiency. We generalize structured pruning with algorithms for network augmentation, pruning, sub-network collapse and removal. In addition, we demonstrate efficient and stable convergence up to 93% sparsity and 95% FLOPs reduction without loss of inference accuracy using with continuous relaxation matching or exceeding the state of the art for all structured pruning methods. The resulting CNN executes efficiently on GPU hardware without computationally expensive sparse matrix operations. We achieve this with routine automatable operations on classification and segmentation problems using CIFAR-10, ImageNet, and CityScapes datasets with the ResNet and U-NET network architectures.

James Herman, Jonathan Francis, Siddha Ganju et al.

Existing research on autonomous driving primarily focuses on urban driving, which is insufficient for characterising the complex driving behaviour underlying high-speed racing. At the same time, existing racing simulation frameworks struggle in capturing realism, with respect to visual rendering, vehicular dynamics, and task objectives, inhibiting the transfer of learning agents to real-world contexts. We introduce a new environment, where agents Learn-to-Race (L2R) in simulated competition-style racing, using multimodal information--from virtual cameras to a comprehensive array of inertial measurement sensors. Our environment, which includes a simulator and an interfacing training framework, accurately models vehicle dynamics and racing conditions. In this paper, we release the Arrival simulator for autonomous racing. Next, we propose the L2R task with challenging metrics, inspired by learning-to-drive challenges, Formula-style racing, and multimodal trajectory prediction for autonomous driving. Additionally, we provide the L2R framework suite, facilitating simulated racing on high-precision models of real-world tracks. Finally, we provide an official L2R task dataset of expert demonstrations, as well as a series of baseline experiments and reference implementations. We make all code available: https://github.com/learn-to-race/l2r.

Anirudh Koul, Siddha Ganju, Meher Kasam et al.

Traditionally, academic labs conduct open-ended research with the primary focus on discoveries with long-term value, rather than direct products that can be deployed in the real world. On the other hand, research in the industry is driven by its expected commercial return on investment, and hence focuses on a real world product with short-term timelines. In both cases, opportunity is selective, often available to researchers with advanced educational backgrounds. Research often happens behind closed doors and may be kept confidential until either its publication or product release, exacerbating the problem of AI reproducibility and slowing down future research by others in the field. As many research organizations tend to exclusively focus on specific areas, opportunities for interdisciplinary research reduce. Undertaking long-term bold research in unexplored fields with non-commercial yet great public value is hard due to factors including the high upfront risk, budgetary constraints, and a lack of availability of data and experts in niche fields. Only a few companies or well-funded research labs can afford to do such long-term research. With research organizations focused on an exploding array of fields and resources spread thin, opportunities for the maturation of interdisciplinary research reduce. Apart from these exigencies, there is also a need to engage citizen scientists through open-source contributors to play an active part in the research dialogue. We present a short case study of SpaceML, an extension of the Frontier Development Lab, an AI accelerator for NASA. SpaceML distributes open-source research and invites volunteer citizen scientists to partake in development and deployment of high social value products at the intersection of space and AI.

Siddha Ganju, Anirudh Koul, Alexander Lavin et al.

Research with AI and ML technologies lives in a variety of settings with often asynchronous goals and timelines: academic labs and government organizations pursue open-ended research focusing on discoveries with long-term value, while research in industry is driven by commercial pursuits and hence focuses on short-term timelines and return on investment. The journey from research to product is often tacit or ad hoc, resulting in technology transition failures, further exacerbated when research and development is interorganizational and interdisciplinary. Even more, much of the ability to produce results remains locked in the private repositories and know-how of the individual researcher, slowing the impact on future research by others and contributing to the ML community's challenges in reproducibility. With research organizations focused on an exploding array of fields, opportunities for the handover and maturation of interdisciplinary research reduce. With these tensions, we see an emerging need to measure the correctness, impact, and relevance of research during its development to enable better collaboration, improved reproducibility, faster progress, and more trusted outcomes. We perform a case study of the Frontier Development Lab (FDL), an AI accelerator under a public-private partnership from NASA and ESA. FDL research follows principled practices that are grounded in responsible development, conduct, and dissemination of AI research, enabling FDL to churn successful interdisciplinary and interorganizational research projects, measured through NASA's Technology Readiness Levels. We also take a look at the SpaceML Open Source Research Program, which helps accelerate and transition FDL's research to deployable projects with wide spread adoption amongst citizen scientists.

Suhas Kotha, Anirudh Koul, Siddha Ganju et al.

A common class of problems in remote sensing is scene classification, a fundamentally important task for natural hazards identification, geographic image retrieval, and environment monitoring. Recent developments in this field rely label-dependent supervised learning techniques which is antithetical to the 35 petabytes of unlabelled satellite imagery in NASA GIBS. To solve this problem, we establish CELESTIAL-a self-supervised learning pipeline for effectively leveraging sparsely-labeled satellite imagery. This pipeline successfully adapts SimCLR, an algorithm that first learns image representations on unlabelled data and then fine-tunes this knowledge on the provided labels. Our results show CELESTIAL requires only a third of the labels that the supervised method needs to attain the same accuracy on an experimental dataset. The first unsupervised tier can enable applications such as reverse image search for NASA Worldview (i.e. searching similar atmospheric phenomenon over years of unlabelled data with minimal samples) and the second supervised tier can lower the necessity of expensive data annotation significantly. In the future, we hope we can generalize the CELESTIAL pipeline to other data types, algorithms, and applications.

Abhigya Sodani, Michael Levy, Anirudh Koul et al.

Researchers often spend weeks sifting through decades of unlabeled satellite imagery(on NASA Worldview) in order to develop datasets on which they can start conducting research. We developed an interactive, scalable and fast image similarity search engine (which can take one or more images as the query image) that automatically sifts through the unlabeled dataset reducing dataset generation time from weeks to minutes. In this work, we describe key components of the end to end pipeline. Our similarity search system was created to be able to identify similar images from a potentially petabyte scale database that are similar to an input image, and for this we had to break down each query image into its features, which were generated by a classification layer stripped CNN trained in a supervised manner. To store and search these features efficiently, we had to make several scalability improvements. To improve the speed, reduce the storage, and shrink memory requirements for embedding search, we add a fully connected layer to our CNN make all images into a 128 length vector before entering the classification layers. This helped us compress the size of our image features from 2048 (for ResNet, which was initially tried as our featurizer) to 128 for our new custom model. Additionally, we utilize existing approximate nearest neighbor search libraries to significantly speed up embedding search. Our system currently searches over our entire database of images at 5 seconds per query on a single virtual machine in the cloud. In the future, we would like to incorporate a SimCLR based featurizing model which could be trained without any labelling by a human (since the classification aspect of the model is irrelevant to this use case).

Sayak Paul, Siddha Ganju

Floods wreak havoc throughout the world, causing billions of dollars in damages, and uprooting communities, ecosystems and economies. The NASA Impact Flood Detection competition tasked participants with predicting flooded pixels after training with synthetic aperture radar (SAR) images in a supervised setting. We propose a semi-supervised learning pseudo-labeling scheme that derives confidence estimates from U-Net ensembles, progressively improving accuracy. Concretely, we use a cyclical approach involving multiple stages (1) training an ensemble model of multiple U-Net architectures with the provided high confidence hand-labeled data and, generated pseudo labels or low confidence labels on the entire unlabeled test dataset, and then, (2) filter out quality generated labels and, (3) combine the generated labels with the previously available high confidence hand-labeled dataset. This assimilated dataset is used for the next round of training ensemble models and the cyclical process is repeated until the performance improvement plateaus. We post process our results with Conditional Random Fields. Our approach sets a new state-of-the-art on the Sentinel-1 dataset with 0.7654 IoU, an impressive improvement over the 0.60 IoU baseline. Our method, which we release with all the code and models, can also be used as an open science benchmark for the Sentinel-1 dataset.

Deep Patel, Erin Gao, Anirudh Koul et al.

Data imbalance is a ubiquitous problem in machine learning. In large scale collected and annotated datasets, data imbalance is either mitigated manually by undersampling frequent classes and oversampling rare classes, or planned for with imputation and augmentation techniques. In both cases balancing data requires labels. In other words, only annotated data can be balanced. Collecting fully annotated datasets is challenging, especially for large scale satellite systems such as the unlabeled NASA's 35 PB Earth Imagery dataset. Although the NASA Earth Imagery dataset is unlabeled, there are implicit properties of the data source that we can rely on to hypothesize about its imbalance, such as distribution of land and water in the case of the Earth's imagery. We present a new iterative method to balance unlabeled data. Our method utilizes image embeddings as a proxy for image labels that can be used to balance data, and ultimately when trained increases overall accuracy.

Sarah Chen, Esther Cao, Anirudh Koul et al.

Due to the nature of their pathways, NASA Terra and NASA Aqua satellites capture imagery containing swath gaps, which are areas of no data. Swath gaps can overlap the region of interest (ROI) completely, often rendering the entire imagery unusable by Machine Learning (ML) models. This problem is further exacerbated when the ROI rarely occurs (e.g. a hurricane) and, on occurrence, is partially overlapped with a swath gap. With annotated data as supervision, a model can learn to differentiate between the area of focus and the swath gap. However, annotation is expensive and currently the vast majority of existing data is unannotated. Hence, we propose an augmentation technique that considerably removes the existence of swath gaps in order to allow CNNs to focus on the ROI, and thus successfully use data with swath gaps for training. We experiment on the UC Merced Land Use Dataset, where we add swath gaps through empty polygons (up to 20 percent areas) and then apply augmentation techniques to fill the swath gaps. We compare the model trained with our augmentation techniques on the swath gap-filled data with the model trained on the original swath gap-less data and note highly augmented performance. Additionally, we perform a qualitative analysis using activation maps that visualizes the effectiveness of our trained network in not paying attention to the swath gaps. We also evaluate our results with a human baseline and show that, in certain cases, the filled swath gaps look so realistic that even a human evaluator did not distinguish between original satellite images and swath gap-filled images. Since this method is aimed at unlabeled data, it is widely generalizable and impactful for large scale unannotated datasets from various space data domains.

Athanasios Vlontzos, Gabriel Sutherland, Siddha Ganju et al.

Future short or long-term space missions require a new generation of monitoring and diagnostic systems due to communication impasses as well as limitations in specialized crew and equipment. Machine learning supported diagnostic systems present a viable solution for medical and technical applications. We discuss challenges and applicability of such systems in light of upcoming missions and outline an example use case for a next-generation medical diagnostic system for future space operations. Additionally, we present approach recommendations and constraints for the successful generation and use of machine learning models aboard a spacecraft.

Panagiotis Tigas, Téo Bloch, Vishal Upendran et al.

Modeling and forecasting the solar wind-driven global magnetic field perturbations is an open challenge. Current approaches depend on simulations of computationally demanding models like the Magnetohydrodynamics (MHD) model or sampling spatially and temporally through sparse ground-based stations (SuperMAG). In this paper, we develop a Deep Learning model that forecasts in Spherical Harmonics space 2, replacing reliance on MHD models and providing global coverage at one minute cadence, improving over the current state-of-the-art which relies on feature engineering. We evaluate the performance in SuperMAG dataset (improved by 14.53%) and MHD simulations (improved by 24.35%). Additionally, we evaluate the extrapolation performance of the spherical harmonics reconstruction based on sparse ground-based stations (SuperMAG), showing that spherical harmonics can reliably reconstruct the global magnetic field as evaluated on MHD simulation.

Alexander Lavin, Ciarán M. Gilligan-Lee, Alessya Visnjic et al.

The development and deployment of machine learning (ML) systems can be executed easily with modern tools, but the process is typically rushed and means-to-an-end. The lack of diligence can lead to technical debt, scope creep and misaligned objectives, model misuse and failures, and expensive consequences. Engineering systems, on the other hand, follow well-defined processes and testing standards to streamline development for high-quality, reliable results. The extreme is spacecraft systems, where mission critical measures and robustness are ingrained in the development process. Drawing on experience in both spacecraft engineering and ML (from research through product across domain areas), we have developed a proven systems engineering approach for machine learning development and deployment. Our "Machine Learning Technology Readiness Levels" (MLTRL) framework defines a principled process to ensure robust, reliable, and responsible systems while being streamlined for ML workflows, including key distinctions from traditional software engineering. Even more, MLTRL defines a lingua franca for people across teams and organizations to work collaboratively on artificial intelligence and machine learning technologies. Here we describe the framework and elucidate it with several real world use-cases of developing ML methods from basic research through productization and deployment, in areas such as medical diagnostics, consumer computer vision, satellite imagery, and particle physics.

Siddha Ganju, Olga Russakovsky, Abhinav Gupta

Collecting fully annotated image datasets is challenging and expensive. Many types of weak supervision have been explored: weak manual annotations, web search results, temporal continuity, ambient sound and others. We focus on one particular unexplored mode: visual questions that are asked about images. The key observation that inspires our work is that the question itself provides useful information about the image (even without the answer being available). For instance, the question "what is the breed of the dog?" informs the AI that the animal in the scene is a dog and that there is only one dog present. We make three contributions: (1) providing an extensive qualitative and quantitative analysis of the information contained in human visual questions, (2) proposing two simple but surprisingly effective modifications to the standard visual question answering models that allow them to make use of weak supervision in the form of unanswered questions associated with images and (3) demonstrating that a simple data augmentation strategy inspired by our insights results in a 7.1% improvement on the standard VQA benchmark.