Neural System

Drone Herd Counting

2 videos · Inventory of cattle and sheep on pasture from drone flyovers

Try with my video

About this solution

Automatic animal counting with fixed cameras or drones: cattle passing through a gate, chute or alley (virtual line), herds on pasture or in feedlots seen from above, sheep and goats in pens and poultry in barns. The AI identifies each animal's species, separates herding dogs and people, estimates herd size and measures area occupancy — replacing slow, error-prone manual "head counts".

  1. 1Text-prompted detection (OWLv2): the AI looks for "a cow", "a sheep", "a goat", "a horse", "a pig" or "a chicken" in each frame, without specific training — it works with any breed and in drone views.
  2. 2For ground cameras there is a fast mode (YOLO11 trained on COCO) that analyzes more frames per second.
  3. 3Dense herds or overhead footage: the frame is sliced into overlapping tiles (SAHI technique) and each tile is analyzed at higher resolution; detections are merged without duplicating animals.
  4. 4Each animal gets a track (ByteTrack); when it crosses the virtual line at the gate or chute it is counted only once, by species.
  5. 5Estimated herd = highest stable count in the frame (median of 5 frames), useful for pasture and feedlot flyovers.
  6. 6Herding dogs and people are marked but not included in the count.
  7. 7Occupancy = share of the image covered by animals; with a stocking limit configured, it raises an alert with a snapshot of the moment.

Benefits

  • Eliminates manual "head counts", which require stopping handling and go wrong on large lots.
  • Counting by species and direction of passage, with video recording as evidence.
  • A drone flyover becomes a herd inventory in minutes, without rounding animals up in the corral.
  • Text-prompted detection enables new species without training a model.
View solution details

More demos