Built at Accelerant

One picture.
Is this area flooded?

TideWatch reads a photo taken from the air or a satellite tile and says whether the area in it is flooded, with the probability behind the answer. It is a research preview, built to see how far a large vision model can go on catastrophe imagery.

Runs inside Accelerant’s environment. Images are scored and never stored.

From a drone photo to a clear answerExample
A school compound and its fields under brown floodwater, seen from a drone

AIFloodSense, CC BY 4.0 · aerial model

Answer

Flooded

Probability

>99.9%

The probability that the area is flooded. Above 50%, the answer is flooded.

Two views, two models

From a drone or from orbit.

A photo taken at an angle and a tile looking straight down share almost nothing a model should learn from, so each view has its own. Pick the one that matches the picture.

Aerial

Drone and helicopter photos, taken at an angle.

A school compound
Flooded>99.9%A school compound
A field and a track
Not flooded0.1%A field and a track

Satellite

The same Nepal tile, before and after the 2017 monsoon floods.

Before
Not flooded<0.1%Before
After
Flooded>99.9%After

Images held out of training, scored by the running service. Sources: AIFloodSense (CC BY 4.0), AIDER, and xBD satellite imagery by Maxar (CC BY-NC-SA 4.0).

How it performs

Measured on images it never trained on.

Each model was scored once on a test set held back from the start. The harder test leaves a whole dataset or flood event out of training: the figure to expect on a new one.

Aerial · test accuracy

98.3%

3,034 test images · 92.3%–99.8% on data left out of training

Satellite · test accuracy

98.0%

594 test images · 70.0%–94.1% on data left out of training

Left out of training entirely
ModelHeld outImagesAccuracy
AerialThe AIDER dataset6,43398.8%
AerialThe AIFloodSense dataset47099.6%
AerialFloodNet, Hurricane Harvey drone photos2,34397.8%
AerialLADI v2, Civil Air Patrol photos, 2015–20239,96992.3%
AerialRescueNet, Hurricane Michael drone photos (none flooded)4,49499.8%
SatelliteSpaceNet 8 tiles1,37774.2%
SatelliteGermany floods, 202137185.7%
SatelliteHurricane Florence, 201857474.4%
SatelliteHurricane Harvey, 201772691.5%
SatelliteLouisiana floods (Ida), 20211,00670.0%
SatelliteMidwest floods, 201930694.1%
SatelliteNepal floods, 20171,02492.1%
SatelliteThe xBD flood tiles2,63088.3%
Every metric, against flexiclass on the same images →

How it decides

A large model looks.
A small one decides.

  1. 01Read the picture

    Perception Encoder, an open vision model from Meta, reads the image and describes it as 1,536 numbers. It is used as it is, never retrained.

  2. 02Weigh the evidence

    A small classifier, trained on 20,652 labelled images from 7 datasets, turns those numbers into the probability of a flood. Each image is read 4 times and the 4 probabilities averaged: a satellite tile, which has no up, as it is, mirrored both ways and turned a quarter; an aerial photo as it is, mirrored, cropped to its centre and tilted 7°.

  3. 03Answer

    Above 50% the answer is flooded. The probability travels with it, so a borderline picture can go to a person.

Know its limits

A research preview.

For developers

1 request, any number of answers.

Up to 16 images a request, JPEG, PNG or WebP, as base64 or a data: URL; the answers keep their order. The key goes in X-API-Key or as a bearer token. GET /v1/tidewatch/image returns the models and these figures as JSON.

curl -X POST "$HOST/v1/tidewatch/image" \
  -H "X-API-Key: $KEY" -H "Content-Type: application/json" \
  -d '{"model": "aerial", "images": [{"data": "'"$(base64 -i photo.jpg)"'"}]}'

{"model": "aerial", "trained": "2026-11-20", "threshold": 0.5,
 "results": [{"flooded": true, "probability": 0.9996}]}

Start with your own picture

Bring a photo. Get an answer.

Choose an example or upload your own image, pick the view, and see the answer and its probability.