Fig. 2From: DeepSeedling: deep convolutional network and Kalman filter for plant seedling detection and counting in the fieldDetection performance (F1 score) calculated using different number of training images for: a the TAMU2015 dataset, b the UGA2015 dataset, and c the UGA2018 dataset. When \(\text {IOU}_{\text {all}}\) (a more strict metric) was used, increasing trends of model performance were clearly observed by using more training imagesBack to article page