JL
Cambiar a español
Back to projects
Independent/portfolio experience

ALA System

Edge computer-vision and IoT automation designed to protect ducks in aviary farms at night without manual intervention, featuring threshold tracking and model fine-tuning.

Evidence status: Approved written profileObject detection · Multi-object tracking · IoT logs · Zigbee
FIG. 01 · PHYSICAL → EDGE → CLOUD

ALA / ENTRY COUNTING RIG

ALA / ENTRY COUNTING RIGIllustrative entry-counting rig with camera, ducks, vertical gate, local Edge controller and remote review server.00AB
INSIDE
2
OUTSIDE
0
GATE
OPEN
EDGE
TRACKING
SERVER
WAITING FOR DAILY RECORD

StatusREADY

Illustrative operating cycle: a camera and edge controller track directional crossings at a controlled entrance. When the configured group is inside, the local controller closes the gate and sends the day’s reviewed event record to a central server for audit and dataset refinement.

Illustrative demo values. The count shows one confirmed crossing per duck through the threshold; it does not describe the production algorithm.

01

An automated edge-to-server vision and gate control system eliminates nightly manual bird roundups, securing ducks inside before nightfall and capturing field data to retrain models and fine-tune parameters.

02

Aviary duck farms require ensuring every bird is safely enclosed inside before dark to prevent predation and injury. Manual headcounts and gate closing every night are labor-intensive, error-prone, and stressful for the flock. The system must operate autonomously at the edge even if the local network is degraded, while continuously adapting to varying flock sizes, lighting, and weather conditions.

People evaluating edge computer-vision and IoT workflow concepts for controlled physical access.

03

Independent/portfolio work: designed the explanatory system concept and interface.

04

  1. Monitor the aviary shed entrance continuously with an overhead optical camera
  2. Edge AI runs real-time object detection and multi-object tracking to record directional crossings
  3. Compute live inside and outside flock counts directly on the edge controller
  4. Verify all ducks are inside by comparing verified occupancy against farm configuration
  5. Trigger vertical gate closure automatically via local Zigbee actuators to seal the aviary before nightfall
  6. Transmit daily crossing logs, edge metrics, and flagged edge-case video clips to the central server
  7. Curate edge-case imagery on the central server to retrain detection models and automatically fine-tune tracking parameters for edge deployment

05

  • Kept counting, threshold logic, and gate actuation completely autonomous on the local edge controller to guarantee birds are safely locked in regardless of Internet connectivity.
  • Automated the nightly gate closure sequence to eliminate manual farm intervention and human error.
  • Established a feedback pipeline where difficult detections and edge cases are uploaded to central storage to retrain models and automatically fine-tune tracker hyperparameters.