Skip to main content
Heritage Science

Table 5 The mean average precision for the models on the validation set

From: Automated corrosion detection in Oddy test coupons using convolutional neural networks

Model

\(\mathbf {mAP}\)

\(\mathbf {mAP}_{0.5}\)

\(\mathbf {mAP}_{0.75}\)

EffDet - V1

0.554

0.925

0.567

EffDet - V2

0.422

0.692

0.500

  1. These calculations were from the COCO API on GitHub [52]. The subscript numbers represent IoU thresholds, and \(\mathbf {mAP}\) was averaged across IoU thresholds from 0.5 increasing to 0.95 by 0.05
Back to article page