Total

0

Unlabeled

0

True Positives

0

False Positives

0

Cameras Online

0

Cameras Offline

0

Health

…

Dataset Storage (quota)

…

0%

…

System

…

By Camera

By Class

Connect Cameras

Configure Gatherer ACAP to publish detections into this platform over MQTT.

ACAP MQTT Settings

On the camera, open Apps and launch Gatherer.
Go to the MQTT tab.
Enter these values and click Connect.

Address

…

Port

1883

Connection mode

Direct (or Tailscale (SOCKS5) if your camera uses Prudscale/Tailscale userspace networking)

User

(blank) unless you configured auth

Password

(blank) unless you configured auth

TLS

No TLS (default)

Pre-topic

prudencia/dataset

Tip: If you are using Tailscale, set Address to the platform server's Tailscale IP (usually 100.x.y.z).
If your camera is on Tailnet via Prudscale/Tailscale userspace networking, set Connection mode to Tailscale (SOCKS5) in the ACAP UI so it can route to 100.x brokers.
Important: Address must be a bare IP/hostname only (no http:// and no :1883).
Optional: set Name and Location in the ACAP UI so the platform can show human-friendly camera info.

Topic Format

Gatherer ACAP prepends Pre-topic automatically. Your code should publish to:

prudencia/dataset/<camera_serial>/<class_label>

Example: publish <camera_serial>/person and set Pre-topic to prudencia/dataset.

Payload Contract

Single JSON message per detection. Images are base64-encoded JPEGs.

After the first message arrives, the camera will appear under Cameras.

Model Inspector

Upload a YOLOv5 .tflite and the platform will extract the quantization values you must enter in the Gatherer ACAP (Model tab). This avoids running any inspection on the camera.

Upload

TFLite model file

Tip: If you have two exports (ARTPEC-8 per-tensor and ARTPEC-9 per-channel), analyze both and save the values per camera.

Result

Model type

-

Input

-

Output

-

Weights quant

-

YOLO stride / classes

-

Fingerprint (sha256)

-

Values For Gatherer ACAP

Copy these into the camera ACAP: Apps -> Gatherer -> Model.

Analyze a model to see values…

Settings fragment for settings.model (useful if you automate camera config).

{}

Select a detection

Crop

Full Frame

Notes

Export YOLO Dataset

Cameras (optional)

Classes (optional)

Export name (optional)

Generate train/val/test split

Augmentations

Include original image Add crop around bbox Horizontal flip Vertical flip Rotate 90 deg clockwise Brightness +/-15% Contrast +/-15%

Only `true_positive` detections are included. Leave cameras/classes empty to include all.

Review Class Boxes

Select exactly one class above. Review shows true_positive detections for that class (filtered by selected cameras, or all cameras if none are selected).

Class for current image

Recent classes

Crop

Full Frame (drag to draw bbox)

Export History

Cameras

Search Group by Category filter Trigger filter

System

Mass ACAP package deployment (.eap) to selected cameras.

Target cameras

Select one or more cameras.

Deploy ACAP Package

Package (.eap) Address override (optional) Start package after install Force deploy if preflight health is red Package name (for start) License key file (optional) Apply MQTT settings after install

Uses camera HTTP API (`/axis-cgi/applications/upload.cgi`). Set per-camera credentials in the Target cameras list (cogwheel). Optional post-install bootstrap can install a license key file and configure MQTT.

Deploy SLO

-

No deploy metrics yet.

{}

Live Deploy Timeline

[]

Last Result

{}