{"slug":"quakeheroes","name":"42 QuakeHeroes","hackathon":"nasa-space-apps-2024","tagline":"A U-Net deep network on decomposed seismic energy that finds the onset of moonquakes and marsquakes in noisy lander data.","placement":{"status":"track-winner","awards":["Best Use of Technology"],"track":"seismic-detection-solar-system"},"team":{"size":6,"members":[],"countries":["BR"]},"problem":"Planetary landers cannot afford the power to send continuous seismic data home, yet only a fraction of it contains quakes.","solution":"A pipeline that band-pass filters Apollo and Mars InSight ground-velocity data, decomposes it with empirical mode decomposition, computes sliding-window energy per mode into a 2D matrix, and trains a U-Net encoder-decoder against Gaussian labels centred on each event's arrival time. A companion web page shows detected moonquake and marsquake types.","demo_moment":null,"build_style":"software-only","tech":{"stack":["python","u-net","empirical-mode-decomposition","apollo-seismic-data","insight-seismic-data","github-pages"],"ai_patterns":["ml-model"]},"domains":["space-science"],"analysis":{"why_it_stood_out":[{"claim":"Won the award for the most innovative use of technology.","basis":"organizer-stated","source":1},{"claim":"Used a deep neural network to locate seismic events precisely in time-series data, with advanced signal processing to isolate components of non-stationary signals.","basis":"organizer-stated","source":0},{"claim":"Validated on held-out data: 63 lunar events for training, 13 for validation, and both available Mars cases used only to check that a Moon-trained model generalizes.","basis":"team-stated","source":2},{"claim":"The write-up explains each step and why (filtering alone failed on low signal-to-noise, so they added decomposition), which makes the depth visible to expert judges.","basis":"curator-inference"}],"mechanisms":["visible-engineering-depth","trend-wave-timing","complete-submission-package"],"judging_lens":["technical-depth","scientific-validity","challenge-relevance"],"lessons":["For a technology award, show the pipeline and the reason for each step, not only the result.","State the validation split, especially when the target environment has almost no data.","LATAM precedent: a team from Brazil won Best Use of Technology."],"confidence":"high"},"links":{"project_page":"https://www.spaceappschallenge.org/nasa-space-apps-2024/find-a-team/42-quakeheroes/?tab=project","video":"https://www.youtube.com/watch?v=thxLntBLF1I"},"sources":[{"url":"https://www.nasa.gov/learning-resources/stem-engagement-at-nasa/nasa-international-space-apps-challenge-announces-2024-global-winners/","kind":"official-results","retrieved_at":"2026-09-14"},{"url":"https://www.spaceappschallenge.org/nasa-space-apps-2024/awards/","kind":"official-event-page","retrieved_at":"2026-09-14","note":"Award definition."},{"url":"https://www.spaceappschallenge.org/nasa-space-apps-2024/find-a-team/42-quakeheroes/?tab=project","kind":"project-page","retrieved_at":"2026-09-14","note":"Team-written method, training and validation split, links to notebook repository and web page."}]}