OCAPI
OCAPI is a web platform for annotating camera-trap images, published by the French company TerrOïko. It pairs automatic species recognition with human validation, and targets ecologists, protected-area managers and environmental consultancies handling large photo volumes.
What is OCAPI?
OCAPI is a web platform for annotating camera-trap data with the help of artificial intelligence. It is published by TerrOïko, a French company founded in 2012 by two ecology PhDs and based in Sorèze, in the Tarn. The problem it attacks will be familiar to anyone running camera traps: a season in the field produces tens of thousands of photographs, most of them empty, and labelling them by hand is the bottleneck. TerrOïko's own order of magnitude puts manual annotation at roughly 500 images an hour, so 5,000 images cost about a full working day.
The platform runs uploaded images through two chained models. MegaDetector first sorts what is in the frame — animal, human or vehicle — and sets aside empty shots. FaunIA, TerrOïko's own classifier, then identifies the species among twelve species or species groups: badger, caprine, red deer, cat, roe deer, dog, lagomorph, small mammal, mustelid, bird, red fox and wild boar, alongside the human and vehicle classes. Average precision is advertised at 97%, and the product page publishes recall class by class. What comes out is a pre-annotation — a bounding box plus a taxonomic label — that the user then validates, edits or deletes. The publisher calls the approach semi-automatic and presents the retained human check as a safeguard, claiming an eightfold acceleration of annotation projects on the strength of Fennell et al. 2022.
Data is organised on three levels: an image belongs to a series, which stands for one collection session at one sensor in one place, and series are grouped into projects. Each series carries its own confidentiality setting, public or private, and its own opt-in for training reuse. Projects can be shared with collaborators, and several annotators can be assigned to the same image so that their annotations are combined. Species names follow TaxRef, the reference of the French National Museum of Natural History, updated once a year.
The boundaries deserve stating up front. OCAPI is browser-only; there is no mobile application and no API. Interface and user documentation exist in French alone. And the site itself gives two different expansions of the OCAPI acronym, one on the product page and another in the documentation.
What it does
- Upload series of camera-trap images into a browser workspace
- Let the AI pre-annotate each photo: human, vehicle, animal or empty frame, then species
- Validate, correct or delete every AI label and bounding box
- Split an annotation project across several collaborators
- Filter the images left to annotate by project, series or detected species
- Track progress through per-series statistics
- Export images and annotations, with labels in JSON and images sorted by category
When to use OCAPI / When not to
A quick filter to help you decide if OCAPI is the right fit.
When to use OCAPI
- Ecology teams facing large backlogs of camera-trap images to label
- Annotation projects shared between several contributors, on Pack 1 or Pack 2
- Studies of population size, animal behaviour and species characteristics built on camera-trap data
- French and European users, since species labelling follows the national TaxRef reference
- Organisations that want AI to do the first pass without giving up human validation
When not to use OCAPI
- Anyone working with images that do not come from camera traps: the platform handles field-sensor photographs only
- Teams whose target taxa fall outside the twelve FaunIA classes, where annotation goes back to being fully manual
- Bird-focused surveys: recall on the bird class is only 0.52, so around half the individuals are missed by the automatic pass
- Non-French speakers: the interface and the whole user documentation exist in French only
- Anyone needing an API, an automated pipeline or per-image GPS positioning, none of which is available; opening an account also requires a company name and registration number
How to use OCAPI
A typical end-to-end flow, from setup to results.
- Request an account on the platform's registration page, giving your name, e-mail, company details, company registration number and the pack you want
- Wait for the request to be processed: TerrOïko answers within two working days and confirms by e-mail
- Create a series from "Gestion des séries" then "Nouvelle série", setting owner, name, confidentiality, training authorisation, AI model and administrative location (country, region, department)
- Select your files and start the upload; EXIF DateTimeOriginal timestamps are summarised for information
- On Pack 1 or Pack 2, create a project and add your series and your collaborators to it
- Invite a collaborator from "Gestion des contacts" then "Créer une invitation", and send them the join code
- Filter the images left to annotate by project, series or detected species, then apply the filter
- Annotate: search for the species, draw a bounding box tight on the animal, switch between creation and edition modes, and skip an image when you are unsure
- Identify at the finest taxonomic level you are certain of, falling back to genus, family or class otherwise, and use zoom and contrast on dark images
- Follow the per-series statistics — images uploaded, annotated, completed, percentage done — then export from the series detail page
Pros & Cons
Pros
- Performance is published class by class, precision and recall alike, which is rare enough to let a buyer judge the model instead of trusting a headline figure
- Human validation is kept throughout: the AI proposes, the annotator decides
- Reuse of data for model training is an explicit opt-in taken series by series, not a blanket clause
- Species labelling follows TaxRef, the French national reference, rather than an in-house nomenclature
- The free Discovery pack is a permanent tier, so the tool can be tried without commitment or countdown
- Public user documentation is detailed and includes annotation best practices
- The time saving is quantified and sourced, an eightfold acceleration credited to Fennell et al. 2022
Cons
- No terms of service and no privacy policy are published for the platform; the only legal text is a legal-notice page
- No data processing agreement, no list of sub-processors and no indication of where user data is hosted
- Neither a series nor a project can be deleted, and removing a user account has to go through support
- Interface and user documentation are in French only, while the publisher's showcase website is bilingual
- No API and no mobile application: everything happens in the browser, and export needs Chrome 86+, Edge 86+ or Opera 72+
- Recall falls on several classes — 0.52 for birds, 0.67 for cats, 0.74 for caprines, red deer and humans — and only twelve species classes are covered, so anything else stays manual
- Opening an account is neither instant nor open to individuals: requests take two working days and call for a company name and registration number
Pricing & Plans
A permanently free tier is available: the Discovery pack costs nothing and is a standing plan rather than a time-limited trial. Paid access starts at 400 EUR excluding tax per year for Pack 1, and rises to 950 EUR excluding tax per year for Pack 2. All published amounts are annual and exclusive of tax; no monthly plan and no usage-based pricing is advertised. A tailored quote can be requested from the publisher for needs the three packs do not cover.
- Discovery pack — free — unlimited human
- vehicle
- animal and empty-frame classification
- species classification capped at 100 images
- 100 MB of storage
- export of annotations only
- MegaDetector alone
- with neither projects nor collaboration
- Pack 1 — 400 EUR excluding tax per year — unlimited human
- vehicle
- animal and empty-frame classification
- species classification up to 10
- 000 images
- 10 GB of storage
- export of annotations and images
- MegaDetector plus FaunIA
- Pack 2 — 950 EUR excluding tax per year — unlimited human
- vehicle
- animal and empty-frame classification
- species classification up to 25
- 000 images
- 25 GB of storage
- export of annotations and images
- MegaDetector plus FaunIA
- Tailored offer — price on request — a custom quote can be obtained from the publisher
- with no public rate card
Data, GDPR & hosting
A consolidated view of how OCAPI handles your data.
GDPR overview
GDPR is acknowledged, but only in a legal notice. The page cites European Regulation 2016/679 and the French Data Protection Act of 6 January 1978, and states that users hold rights of access, rectification and erasure. A dedicated address, info.rgpd@terroiko.fr, is published for data-protection questions. The publisher is established in Sorèze, France, so no Article 27 representative is required. The same page adds that TerrOïko collects no personal data on its site — a statement about the showcase website, not about the OCAPI platform. Nothing further is published: the annotation platform has no privacy policy and no terms of service of its own, no data processing agreement is offered, no sub-processor is listed, and no retention period is given. The commitment is genuine but thinly documented.
Who owns the data?
The documentation treats uploaded material as the user's own. Every series carries an explicit Owner field naming the data owner, and its creator sets a confidentiality level: public series can be seen and annotated by any user, private ones only by the creator and the collaborators invited. Reuse of the images by the publisher to train its models is a separate opt-in checkbox, ticked series by series. Beyond that, ownership is not formalised: TerrOïko publishes no terms of service and no privacy policy for the platform, only a legal-notice page, so no contract states what rights the publisher holds over customer images and annotations.
Reuse rights
Uploaded images go through two successive models: MegaDetector first, to separate animals, humans, vehicles and empty frames, then FaunIA to identify the species. That processing is the service itself. Any other reuse requires explicit permission: when creating a series the user may tick "I authorise TerrOïko to use my data to improve the species classification models", in which case the images are used in the next training run of the AI model; left unticked, the series stays out of training. Users can export their own material — images and annotations on the paid packs, annotations alone on the free one — and no published document restricts what they then do with it. One caution: the legal notice's statement that TerrOïko collects no personal data concerns the publisher's showcase website, not the annotation platform.
Data retention & training
Hosting summary
Where OCAPI stores user data is not documented. Neither the product page, nor the user documentation, nor the legal notice says in which country or on which infrastructure uploaded images and annotations are held. The only hosting information published concerns the publisher's showcase website, which the legal notice declares hosted by OVH, 2 rue Kellermann, 59100 Roubaix, France — a statement about the website, not about the annotation platform, and one that cannot be carried across to it. A network check adds a caution rather than an answer: terroiko.fr resolves to 80.11.154.7, an address located in France on Orange's AS3215, which does not match the OVH declaration and confirms that nothing can be deduced about where customer data sits. What can be said is that the publisher, TerrOïko, is a company established in Sorèze, France, and so operates under French and EU law. No data processing agreement, no sub-processor list and no choice of hosting region are offered. Anyone with a hosting requirement should obtain an answer in writing before uploading anything.
Things to keep in mind
Risks and trade-offs to weigh before adopting OCAPI.
- The platform has no privacy policy and no terms of service: reading the legal notice will not tell you what becomes of your images once they are uploaded
- The statement that TerrOïko collects no personal data covers the publisher's showcase website, not the annotation application — do not read it as a commitment about your own data
- Where user data is hosted is never published; the host named in the legal notice covers the website, and the site's DNS resolution points to a different network altogether, so nothing can be inferred
- Neither a series nor a project can be deleted, so a mistake in how data is organised is permanent
- Permission to reuse your images for AI training is given at series creation, an easy decision point to click past without noticing
- The user documentation has not been updated since 15 February 2024: check that the packs, quotas and species classes described still hold, and remember that the published prices exclude tax
- The tool is built around TaxRef and European wildlife classes; no country restriction is announced, but check species coverage before using it outside that perimeter
Setup & Integrations
Technical difficulty
Low on the technical side. OCAPI is a web application: nothing to install, nothing to configure, and the documentation walks through series creation, upload, projects, collaborators, annotation and export step by step. What slows a start is administrative rather than technical — an account has to be requested and is granted within two working days, and the form asks for a company name and registration number. Two details to plan for: exporting requires Chrome 86+, Edge 86+ or Opera 72+, and adding GPS coordinates means editing EXIF fields beforehand with an outside tool such as ExifTool.
Deployment
Supported languages
Behind OCAPI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is OCAPI?
Which species does it recognise?
How reliable is the automatic recognition?
How much does it cost?
How do I open an account?
Can several people work on the same data?
Will my images be used to train the AI?
What does an export look like?
Is there an API?
Can I delete a series or a project?
Should you pick OCAPI?
OCAPI is a narrow tool, and that is its strength. It does one thing — turning a backlog of camera-trap photographs into labelled data — and it does it in a way that can be audited. TerrOïko publishes precision and recall class by class instead of a single headline number, keeps a human in the loop on every annotation, makes the reuse of images for model training an explicit choice taken series by series, and offers a free pack that is a permanent tier rather than a countdown. For a French or European ecology team, TaxRef labelling and the twelve species classes covered make it usable straight away; outside that geographical and taxonomic perimeter the case is much weaker.
The reservations are documentary rather than technical. The platform has no terms of service and no privacy policy of its own: the only legal text is a legal-notice page, and it says nothing about where uploaded images are stored or how long they are kept. No data processing agreement is offered and no sub-processor is named. Neither a series nor a project can be deleted once created. Interface and documentation exist in French only, that documentation has not been revised since February 2024, and there is no API to wire OCAPI into an existing pipeline.
Pricing is readable and modest — free, then 400 EUR excluding tax a year, then 950 — and reasonable against the time saving claimed. Before committing, put three questions to the publisher in writing: where the images are hosted, how long they are retained, and how deletion actually works. Ask as well whether the packs, quotas and species classes described in the documentation are still current. Most of the decision rests on those answers.
- Choosing a selection results in a full page refresh.
- Opens in a new window.