Real demo recording · English narration and captions · Fictional data
THE PROBLEM
A useful visual count must let reviewers check for missing or duplicate objects before accepting inventory.
THE WORKFLOW
Local YOLO26s detects plants in an original nursery scene. Review nine visible units, add or remove boxes, correct labels and save the count with its evidence.
01Image and category
02Local instance detection
03Correction and approval
04Exportable count and evidence
RECORDED EVIDENCE
Results, with context.
Results on synthetic data
Check
Cases
Correct count on the scene
1/1
Dataset
original-controlled-fixtures-v1
Completed cases
1/1
Median / p95 per case
11.238 s / 11.238 s
Evaluation record (UTC)
2026-10-08
Source: the project’s evaluation record, using local inference. Recorded hardware: NVIDIA GeForce RTX 4070 Laptop GPU / 8 GB. Latency includes the full HTTP case and may include initial model loading. These measurements have not been rerun on the machine serving this website.
A visual counting pilot with defined categories, review and export to your inventory process.
What we would define first
Categories, cameras, occlusions, capture conditions, acceptable error and integration format.
Demonstration limits
One original AI-generated image with nine visible plants. The count was checked on that scene; it does not represent an entire warehouse or general accuracy. Generic COCO weights are used.
Evaluated demo
Evaluated demo