Fig. 2
From: HairNet2: deep learning to quantify cotton leaf hairiness, a complex genetic and environmental trait
Architecture of the HairNet2 Model. Images were captured using the Image Acquisition (A) protocol described in [8]. The Feature Extraction Module (B) processes input images to derive essential visual features that capture the intricacies of leaf surfaces. The resulting distilled features are fed into the Segmentation Module (C), which differentiates trichomes from cotton leaf surface and produces a segmentation mask. The Outputs (D) consist of two parts: the predicted segmentation mask that highlights trichomes and a Leaf Trichome Score (LTS) quantifying the leaf hairiness and calculated from the segmentation mask