Weakly and semi-supervised detection, segmentation and tracking of table grapes with limited and noisy data

Thomas A. Ciarfuglia, Ionut Marian Motoi, Leonardo Saraceni, Mulham Fawakherji, Alberto Sanfeliu, Daniele Nardi

Abstract: Highlights•A pseudo-labelling system for detection and segmentation is proposed.•The proposed system is low cost and applicable to a wide range of agricultural tasks.•The generated labels consistently improve transfer learning effectiveness.•Experiments show that intra-species covariate shift is significantly reduced.