POPs: Pluck Out the Petals

In this work, we present a method for precisely segmenting individual petals within flower images. Petal segmentation is challenging due to its intricate structures and substantial variations among petals across diverse flower species. Moreover, frequent occlusions and the uniformity in appearance and shape of petals within a single flower add complexity to the segmentation process, making it error-prone. We introduce a generalized approach to segment petal instances for any flower species. We first detect contours for the outermost set of petals and segment them out from the Corolla. We iteratively repeat this process until we reach the floral centre. To this end, we developed a robust contour detection technique and employed a prompt-based segmentation model for petal segmentation. Our results demonstrate the effectiveness of prompts generated by our approach in facilitating efficient petal segmentation. Additionally, we have developed a manual supervision tool, FloraSeg, tailored for segmenting petals in cases where flowers exhibit many petals with complex geometries. Our experiments demonstrate that the fully automatic segmentation method achieves results comparable to those obtained through manual human annotation. Addition-ally, incorporating a few manual cues from the FloraSeg tool into our approach consistently yields precise petal segmentation, aligning closely with manually annotated outcomes. We present results from a wide range of flower species to emphasize the effectiveness of the proposed approach in the petal segmentation task. This work has significant applications in developmental biology, particularly in advancing our understanding of flower development and conducting structural phenotyping of flowers.