Labels for Emergency Response Imagery from Hurricane Florence, Hurricane Michael, and Hurricane Isaias
Bibliographic record
Abstract
The csv files contain human-generated labels for Emergency Response Imagery collected by US National Oceanic and Atmospheric Administration (NOAA) after Hurricane Florence (2018), Hurricane Michael (2018) and Hurricane Isaias (2020). All authors contributed to labeling the imagery. All labeling was done with an open-source labeling tool (Rafique et al., 2020). All csv files provide the userID (the ID of the anonymous labeler), the NOAA flight, the NOAA image, and 6 labels — allWater (if the image was all water), devType (if the image had buildings/development), washoverType (if the image had washover deposits), dmgType (if the image showed damage to built environment), impactType (if the labeler could identify the coastal impact, using the Storm Impact Scale from Sallenger, 2000), and terrainType (the type of physical environment). Images labeled here correspond to 3 NOAA flights — Florence 20180917a , Michael 20181011a, Isaias 20200804a. These images can be downloaded directly from NOAA (https://storms.ngs.noaa.gov/) or using Moretz et al. (2020a, 2020b). There are three csv files: ReleaseData_v3.csv has 6200 labels for 1500 images. These labels were generated by trained coastal scientists. ReleaseDataQuads.csv has 400 labels for 100 images. These labels were generated by trained coastal scientists. The images labeled in this set correspond to original NOAA images that have been split into quadrants. Splitting images was done with ImageMagick. The command used to split the images was: `magick mogrify -crop 2x2@ +repage -path ../quadrants *.jpg` The naming convention corresponds to the image quarter — the *-0.jpg is upper left, *-1.jpg is upper right, *-2.jpg is lower left, and *-3.jpg is the lower right. ReleaseDataNCE.csv has 400 labels for 100 images. These images were labeled by non-coastal scientists. Note that the 100 images were also labeled by coastal scientists — those labels can be found in ReleaseData_v3.csv. There is another companion dataset to this, with slightly different labels (Goldstein et al., 2020)
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.256 | 0.164 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".