CSI Ocean is a free mobile app for reporting microplastic observations in water, air, and snow. Snap a photo, capture the science, and contribute to a global, open dataset that researchers and communities can act on.
CSI Ocean turns your phone into a microplastic field kit. Open the app, photograph a sample of water, air, or snow, and the app's machine-learning model marks suspected microplastic particles with bounding boxes and confidence scores.
You add the local conditions, depth, sample volume, sampling method, location, and submit. The observation lands on a global map that anyone can browse. Researchers get a structured, reproducible dataset. Educators get a real-world experiment their classes can run anywhere. Citizens get a way to see, document, and discuss what their environment is carrying.
Free. No ads. No tracking. No premium tier.
Microplastic pollution is one of the most pervasive environmental challenges of this century, and one of the hardest to measure. Particles smaller than five millimetres travel in ocean currents, blow on the wind, and fall with the snow. They have been found from the Mariana Trench to the summit of Mount Everest.
But the data on where, how, and how much is fragmented. Field samples sit in spreadsheets that never reach the public. Citizen-led efforts use inconsistent methods. Volunteers want to help, but don't know what counts as a useful observation, what to record, or where their photos should go.
CSI Ocean closes that gap. It encodes the methodology of trained microplastics researchers (sample type, environment, depth, volume, equipment, contamination protocols, filtering method) into a guided wizard that anyone can follow. It uses machine-learning detection to flag candidate particles in seconds. It stores everything in one open, geo-located dataset that researchers can query and the public can explore.
A clipboard, a microscope, a lab notebook, and a desktop database used to be the cost of entry. Now a curious person with a phone can do it on a beach, in a park, or in their own backyard.
Standardised capture, structured metadata, and detection results stored alongside every image. Export observations as PDFs for fieldwork records, or use the dataset as a feed into your own analysis pipeline.
See the data modelarrow_forwardA real-world citizen-science project that runs on phones students already own. Use it for a single field trip or a semester-long monitoring programme. The taxonomies read like a textbook chapter your class can fill in.
Run with your classarrow_forwardRun a project, invite your community to follow it, and watch contributions roll in. Each observation is geo-pinned, timestamped, and tied to a sampling protocol your team can vouch for. A built-in leaderboard recognises your most active contributors.
Find or start a projectarrow_forwardYou don't need a lab coat. You need a phone, some daylight, and twenty minutes. Onboarding walks you through your first observation. From there, every walk along the shore, every snowfall, every stream becomes a small contribution to a global picture.
See how it worksarrow_forwardBrowse the public map. Filter by sample type, region, or time window. See where contributions are concentrated, what is being detected, and where the gaps are. Every observation is sourced and dated.
Browse the global maparrow_forwardPick water, air, or snow. The capture wizard adapts to the sample type, asking only for the fields that make sense for what you are documenting.
Take up to 30 photos with the camera, or pick existing ones from your library. CSI Ocean sends them to its detection model and marks every suspected microplastic particle with a numbered bounding box and a confidence band: green for high, amber for medium, red for low.
Depth, water type, environment, sample volume, equipment, filter pore size, contamination protocol, weather, time of day. Only the fields relevant to your sample type. A skip-advanced option keeps it light when you are starting out, or fill it all in for research-grade entries.
Tap your location, drop a pin, or enter coordinates manually. The app reverse-geocodes a human-readable address. Offline? No problem. The observation queues locally and syncs the moment you reconnect.
Your observation lands in the global feed and on the explore map. Browse what others are reporting. Filter by sample type. Switch from clustered pins to a heatmap. Tap any observation to see the full breakdown, including detections drawn directly on the photos. Export a polished PDF for your field journal.
Eight feature groups that reflect how the app is actually built, from the wizard down to the secure session. Skim the headlines, or open the bullets when you want the detail.
A multi-step wizard that adapts to the sample type, machine-learning detection on every photo, and per-image validation status.
A clustered pin map and animated heatmap layer, with filters by sample type and time window, plus a posts feed and global search.
A three-tab detail view (Information, Detected, Data Quality), pinch-zoom on bounding boxes, inline location card, and a polished PDF export.
Public projects anyone can create or follow, project-scoped observations, leaderboards, and a member directory with follow and unfollow.
Email-based signup with explicit consent, six-digit verification, secure password handling, and persistent session in the device secure enclave.
Offline-first capture with status badges, cached scientific vocabularies, and resilient sync when connectivity returns.
Standardised vocabularies prefetched from the server, data-quality tags, detection confidence bands, and soft-delete on the server.
No third-party analytics, no crash-reporting SDK, no ad identifiers, no push notifications, no background location. The app talks to its own backend and to a hosted detection endpoint.
The wizard codifies how a microplastics researcher would walk through the work, sample, environment, equipment, protocol, location, so volunteer observations end up structured the way labs need them.
Every observation in the public dataset is browsable, searchable, and pinnable on a global map. There is no paid tier and no premium dataset.
Photos are sent to a hosted machine-learning endpoint that returns annotated detections in seconds. You see suspected particles right there in the wizard.
Offline capture, cached vocabularies, and resilient sync mean the app works on a boat, in a forest, or under a snowfall, just as well as in a lab.
No third-party analytics SDK is integrated. No advertising network is loaded. The app talks to its own backend and to a hosted detection endpoint, and that is it.
Built with native iOS components and Material guidelines, with text contrast checked against WCAG AA at body sizes. Uses haptics, share sheets, and link previews to feel like a first-class platform citizen.
“Citizen science only matters if the data lands in a form labs can use. CSI Ocean encodes the protocol so contributions come back structured, geo-referenced, and ready to compare across sites. That is rare.”
Switch from clustered pins to a heatmap. Filter by sample type. Tap a pin to read the full observation, including detections drawn on the photos. The cinematic below is rendered programmatically with Remotion to give you a feel for the live screen.
5 sample observations · powered by Google Maps in-app
Sample data shown. The live feed connects when the public dataset ships.
Here is exactly what the app collects, what it does not, and how we protect your account. We treat the scientific record as a public good.
Account fields, plus the scientific metadata of every observation you submit.
The list of things this app deliberately does not include.
Five concrete protections that are in place today.
Yes. There is no paid tier, no premium dataset, and no advertising. The project is funded by a research programme so the app can stay free for the people doing the work.
No. The capture wizard is designed for first-time observers. You'll be guided through every step, with optional advanced fields you can skip until you want them. If you have scientific training, the metadata is rich enough for research-grade observations.
The detection model flags suspected microplastic particles and assigns a confidence band: high, medium, or low. It is intended as a screening aid, not a substitute for laboratory confirmation. Researchers can filter by confidence and by data-quality tag (Casual, Needs ID, Research).
Yes. The full capture wizard works offline. Observations queue locally with a visible status indicator and sync automatically when you reconnect. Every scientific lookup the wizard needs (sample types, environments, equipment, contamination protocols) is cached on first launch.
Photos are uploaded to the project's secure cloud storage and stored privately. They appear in the public dataset attached to your observation, with detection bounding boxes drawn over them. If you delete the observation, the photos are removed from public view.
You can export individual observations as polished PDFs from the observation detail screen, including cover page, metadata, and image pages with detection overlays.
By default, every approved observation appears in the public feed and on the explore map. Your username, sample photo, location, sample type, and detection count are public. Your private profile fields (email, phone, date of birth) are not.
Yes. Citation block (DOI, project lead, institution, suggested citation format) will be published with the open-data release.
Not yet. The app currently ships in English only. Additional locales are on the roadmap.
Email support and we'll process the deletion. Your account record and personal fields are removed; observations may remain in the public dataset in anonymised form, depending on the project's data-retention policy.
Use the in-app feedback form, or email the support address.
Observation media is stored on Amazon S3. Database records are stored on the project's primary cloud-hosted database.
Partner agencies and academic institutions will be listed once collaborations are confirmed.
Researchers and the scientific advisory team review samples flagged for follow-up. Model updates are versioned, and every observation records the detection model version used at capture.
Still have a question? Email the team: see the contact details in the footer below.
Free on iOS and Android. Twenty minutes from a curious walk to a peer-reviewable record. Every observation contributes to a shared scientific dataset.
No ads · No tracking · No premium tier
