Your first research project¶

Before you begin¶

• You have an activated TRAPPER account (registered + approved by an admin)
• You have some camera-trap images or videos ready to upload, organized by deployment
• You know the two places you'll be working in: the Expert admin (/admin/, where projects, deployments and classification setup live) and the Citizen Science frontend (a separate app, usually on its own URL/port — see Citizen Science: first observations for how to log into it), where the actual classifying happens. The tutorial tells you when to switch between them.

Steps¶

1. Create a research project¶

In the left sidebar, open Research → Research projects, then click the + icon in the top right of the list page. This opens the Add research project form.

Fill in the form and click Save. The fields that matter most for a first project:

Description, Abstract, Methods and Tags are optional free text. There are also two inline tables at the bottom of the form (Research project roles, Research project collections) — you can leave both empty for now and come back to them later.

Once approved, the project appears under Research → Research projects, listing Name, Acronym, Keywords and your role. You can filter the list or search it.

2. Add locations¶

Two ways to add Locations: manually, or via file upload. Both live under Geomap → Locations in the sidebar.

Manually: open Geomap → Locations, click the + icon, then place the point on the map view.

Bulk via CSV or GPX (recommended for more than a couple of locations): a comma-separated .csv (gzipped .csv.gz is also accepted) with columns:

From the Locations list, click Import locations from CSV/GPX (next to the + icon), pick your file, the target Research project, and the timezone to apply to every imported location (there's also an Ignore DST checkbox for cameras that don't observe daylight saving time). See Import locations & deployments for the full timezone/DST rules.

3. Add deployments¶

A Deployment adds a start/end time to a Location — one camera-out-to-pickup session, or a session between battery/card changes. Manually adding deployments is error-prone; instead, generate the deployments table from the media files themselves with trapper-tools:

trapper-tools helpers template --data-path /path/to/data

This analyzes your media and writes a CSV with the deployment metadata. Import it from Geomap → Deployments, using the Import deployments from CSV button next to the list's + icon.

4. Upload resources¶

See Convert, package & upload with Trapper-Tools for the full ingestion workflow. The short version:

trapper-tools pipeline --data-path /path/to/data --convert --trigger

This converts media, packages it, uploads via the resumable HTTP chunk uploader, and (with --trigger) automatically starts the AI pipeline if one is configured on the target Classification Project.

5. Build a Classificator¶

A Classificator describes the set of attributes annotators fill in — species, observation type, count, and so on. Open Media classification → Classificators and click the + icon.

Check the pre-defined attributes you need, or add custom ones in the custom attribute form.

6. Create a classification project¶

Open Media classification → Classification projects and click the + icon. Set the name, the parent Research Project, the Classificator from step 5, an optional default AI Provider, and the users involved.

For the full field reference (AI models, privacy/blurring, confidence thresholds), see Create a classification project.

7. Attach resources to the project, and build sequences¶

A Classification Project doesn't see your Collection's resources automatically — you link them with a Classification projects collection row. Open Media classification → Classification projects collections, click the + icon, and pick your Project and Collection from the two dropdowns. Save.

Back on that same list page, tick the checkbox next to the row you just created, choose (Re)build sequences for selected collections from the action dropdown, and run it with a time interval (default 5 minutes; 15 minutes is common in practice). This groups the collection's resources into Sequences — bursts of frames close together in time — which is what lets the classify view step you through a burst instead of one frame at a time.

8. Classify a resource¶

This step happens in the Citizen Science frontend, not the Expert admin — open it in a separate tab and log in with the same TRAPPER account (see Log in if you haven't already). Select your project, then open Classification view in the left sidebar.

Pick your Deployment from the Filters panel — a grid of thumbnails appears below, one per resource, with a green corner mark on resources you've already classified.

Click a thumbnail's Classify link to open it. The strip along the top shows every resource in the same Sequence, labelled Seq 1, Seq 2, … — use it (or the Previous/Next resource and Previous/Next sequence buttons) to move through a burst without going back to the grid. Each thumbnail also has a Select toggle: turn it on for several frames at once if they should all get the same classification (e.g. an empty trailing burst), which lets you classify them together in one go instead of one at a time.

On the resource itself, click Select new object and draw a box around each animal/human/vehicle (or Add object (no bbox) for an observation that isn't tied to one drawable object, e.g. a group-level count). Fill in:

Add a row per animal/object when several are present with different attributes, then click Classify to save it as your own classification. If the resource already has an AI prediction you agree with, you can click Approve AI instead of drawing anything yourself — see Your first AI pipeline run for that flow.

9. Approve the classification¶

Clicking Classify in step 8 only saves your classification — it doesn't make it the project's accepted record yet. That happens back in the Expert admin: open Media classification → Classifications, use the Filters panel to pick your Classification project (the per-project approval actions only appear once a single project is selected), then click Approve USER at the top of the list. This approves every USER classification currently matching your filter, in the background via Celery — there's no need to select individual rows first.

(If you classified by clicking Approve AI directly in step 8, that resource is already approved — no extra step needed. The Expert admin's own Approve AI action does the same thing in bulk, for every AI prediction above a confidence threshold you set.)

Verify it worked¶

Open Media classification → Classifications filtered to your project: the resources you classified should now have a FINAL row with Is approved checked. The Classification Project's resource counts (visible on its own page) reflect the same thing.

Next steps¶

• Your first AI pipeline run
• Bounding boxes for ML training data
• Camtrap DP export