Dragoneye | Turn videos into data

Van counts by westbound movement since 7 AM
Left turn	Straight	Right turn
11	57	18

Models that understand the details

Our models extract rich information from your videos, including the characteristics and history of the objects in them.

Model: roadside_assets

#pg #record .s7-record-photo > .h51-sign-photo:not(.h51-sign-noscript){display:none}

Roadside assetObject R-122

Transit lane sign

Condition4 / 5

Visible text: RIGHT LANE · BUSES TAXIS ONLY
Location: 1000 block Mission StBetween 6th & 7th · right sideApprox. GPS 37.7800, -122.4101

Time tracked2.8 seconds09:42:09.4–09:42:12.2

09:42:09.409:42:12.2

09:42:0009:42:0409:42:0809:42:1209:42:16

A working model in minutes

Your coding agent can define and build Dragoneye models that work out of the box, without data labeling or model training.

Describe what you want to detect

Count vehicles by type, including cars, pickup trucks, vans, buses, motorcycles, and trucks.

Your agent writes the model schema

Use Claude Code or Codex with the Dragoneye MCP server, or the agent in our Playground. Define what your model should detect.

Model schema

modelintersection_vehicles

detect

Vehicle

Passenger car
Pickup truck
Van
Bus
Motorcycle
Truck

Dragoneye builds your model

Dragoneye builds your model from the schema. Ready in minutes, with no training data or labeling.

Schema ready Build pending Ready to use

Model to buildrecognize_anything/intersection_vehicles

Not built

Ready in minutes·No training data or labeling

Run the model on your video

Run the model in a few lines with our Python or Node SDK.

SDK reference ↗

import asyncio
from dragoneye import Dragoneye, Video

client = Dragoneye()
result = asyncio.run(client.classification.predict_video(
    media=Video.from_path("intersection.mp4"),
    model_name="recognize_anything/intersection_vehicles",
))
objects = result.objectsimport { Dragoneye } from "dragoneye-node";

const client = new Dragoneye({
  apiKey: process.env.DRAGONEYE_API_KEY
});
const video = await Dragoneye.Video.fromFilePath(
  "intersection.mp4"
);
const result = await client.classification.predictVideo(
  video,
  "recognize_anything/intersection_vehicles"
);
console.log(result.objects);from dragoneye import Dragoneye, Video

client = Dragoneye()
video = Video.from_path("intersection.mp4")
result = await client.classification.predict_video(
    media=video, model_name="intersection_vehicles"
)import { Dragoneye, Video } from "dragoneye-node";

const client = new Dragoneye();
const video = await Video.fromFile("intersection.mp4");
const result = await client.predictVideo(
  video, "intersection_vehicles"
);

intersection.mp4 · example clip

Review the results

2 passenger cars2 vans1 truck

Object	Vehicle type	First seen	Last seen	Confidence
V-001	Van	00:13.6	00:20.0	0.98
V-002	Van	00:17.2	00:19.6	0.97
V-003	Passenger car	00:17.5	00:20.0	0.96
V-004	Truck	00:15.8	00:20.0	0.98
V-005	Passenger car	00:18.6	00:20.0	0.97

V-001

VanV-001

V-002

VanV-002

V-003

Passenger carV-003

Adapt as your needs change

Update your model to extract the new information your use case needs.

Describe what’s changed

Record each vehicle’s exterior color and make.

Update the schema. Dragoneye rebuilds.

Add attributes to the schema, then send the update to Dragoneye.

Model schemaRevised

modelintersection_vehicles

detect

Vehicle

Passenger car
Pickup truck
Van
Bus
Motorcycle
Truck

attributes Added

Exterior color: White, Silver, Blue, …
Make: Ford, Toyota, Mercedes-Benz, …

Schema ready Build pending Ready to use

Model to rebuildrecognize_anything/intersection_vehicles

Not built

Ready in minutes·No training data or labeling

The new data flows in

Run the rebuilt model on the same video. Exterior color and make now appear alongside each tracked vehicle.

intersection.mp4

Object	Vehicle type	First seen	Last seen	Confidence	Exterior color	Make
V-001	Van	00:13.6	00:20.0	0.98	White	Ram
V-002	Van	00:17.2	00:19.6	0.97	Black	Toyota
V-003	Passenger car	00:17.5	00:20.0	0.96	Black	Ford
V-004	Truck	00:15.8	00:20.0	0.98	Light blue	Mack
V-005	Passenger car	00:18.6	00:20.0	0.97	Black	Toyota

Models that adapt across environments

Pretrained on a broad range of visual data, our zero-shot models adapt to new environments and use cases without task-specific labeling or training.

Stand cam 2 — stand A21, catering on stand at 14:17:15

Airport operations

Track ground equipment

Model result

Equipment: Cargo loader
Bag unload: Open

W 34th St & Eighth Ave, elevated camera, 13:47:06

Traffic surveys

Count movements by vehicle type

Model result

Vehicle type: Van
Westbound · straight: 57 vehicles

Creator feed, session 3, feed time 00:31:48

Sponsor exposure

Measure logo exposure

Model result

Brand: OpenAI
On screen so far: 3.80 seconds

Subject detail

Forward dashcam on Mission Street between 6th and 7th Streets, San Francisco

Roadside inventory

Catalog roadside assets

Model result

Asset type: No-stopping sign
Side of road: Left

Subject detail

Easy to integrate into your product

Use our Python or Node SDK to connect model results to your alerts, reports, and workflows.

Video models

Airport operationsground_equipment

Traffic surveysintersection_vehicles

Roadside inventoryroadside_assets

Structured
results

Dragoneye SDK

Use model
results.

Bring detections into
your application.

Model results

In your productLive outputs

Tug overdueStand A21 · 5 min

Alert

Vehicle flow2,417 vehicles

+6.2%

Transit lane signReview queue

97%

OpsAlertStand monitoring

Stand A2114:58

Pushback tug 5 minutes overdue

OverdueExpected by 14:53

Intersection traffic survey

Vehicle count summary

2,417Vehicles since 7 AM

13.3%Heavy vehicle share

Vehicles per 15 minutes 07:45 · 167 vehicles

170850

07:0010:3013:45

RoadReviewInventoryR-118

Extracted type: No-stopping sign
Road side: Right
Condition: 4 / 5

See what’s possible with your footage

Try it on your own videos, or discuss your use case with us.

Try it on
your videos

Use the Playground to create a model and test your footage.

Try it on your videos

Discuss
your use case

Tell us what you need from your videos. We’ll talk through whether Dragoneye can help.

Discuss your use case

Turn videos into data

Turn videos into data

Transit lane sign

Describe what you want to detect

Your agent writes the model schema

Dragoneye builds your model

Run the model on your video

Review the results

Describe what’s changed

Update the schema. Dragoneye rebuilds.

The new data flows in

Airport operations

Traffic surveys

Sponsor exposure

Roadside inventory

Use model results.

Pushback tug 5 minutes overdue

Intersection traffic survey

Try it on your videos

Discuss your use case

Use model
results.

Try it on
your videos

Discuss
your use case