Table 7 Computational burden analysis for four versions of the proposed system.
From: Multipurpose monitoring system for edible insect breeding based on machine learning
x | % of box area | ISM | SSM | LPM | Total time per box | No. of boxes per day | |||
|---|---|---|---|---|---|---|---|---|---|
Model | Time per box | Model | Time per box | Parameters | Time per box | ||||
1 | 100 (192 tiles) | Mask R-CNN | 30.6 s | U-Net | 8.7 s | Proposed all | 186.4 s | 225.7 s | 380 |
2 | 25 (12 tiles) | Mask R-CNN | 2.0 s | U-Net | 0.5 s | Proposed geometric (length, volume, curvature) | 52.4 s | 54.9 s | 1570 |
3 | 25 (12 tiles) | Mask R-CNN | 2.0 s | U-Net | 0.5 s | Basic (area of binary mask) | 0.4 s | 2.9 s | 30,000 |
4 | 25 (12 tiles) | YOLOv5x | 1.1 s | U-Net | 0.5 s | – | – | 1.6 s | 54,000 |
