Table 9 Detection performance of models on parcel box dataset under different backbones.
From: Research on dimension measurement algorithm for parcel boxes in high-speed sorting system
Method | Params (M) | GFLOPs | P (%) | R (%) | F1 (%) | mAP50 (%) |
|---|---|---|---|---|---|---|
Fasternet | 4 | 15.7 | 92.5 | 89.4 | 90.9 | 95.6 |
EfficientViT (CVPR)30 | 3.88 | 14.5 | 95.5 | 89 | 92.1 | 92.9 |
RevCol31 | 3.65 | 15.3 | 97.3 | 87.1 | 91.1 | 93.4 |
Repvit32 | 6.51 | 23.5 | 92.5 | 88 | 90.2 | 93.5 |
Swin-tiny | 29.72 | 83.8 | 97.1 | 89.2 | 93 | 96.7 |
MobilenetV3-S33 | 2.46 | 10.6 | 95.1 | 94.4 | 94.7 | 97.7 |
PVTv2-B034 | 5.16 | 18.7 | 87.9 | 79.4 | 83.4 | 89 |
CSPDarkNet5335 | 7.01 | 22.7 | 96.6 | 94.9 | 95.7 | 98.2 |
ResNet1836 | 13.11 | 40.0 | 86 | 82.3 | 84.1 | 89.4 |
Repvgg-A037 | 10.52 | 35.0 | 89.9 | 89.8 | 89.8 | 95.8 |
efficientnet_b038 | 5.36 | 16.0 | 96.7 | 90.1 | 93.3 | 96.7 |
(ES-ViT)Ours | 2.68 | 11.7 | 99.3 | 98.4 | 98.8 | 99.5 |
