The future of agriculture is intertwined with automation. Accurate fruit detection, yield estimation, and harvest time prediction are crucial for efficient supply chain management by optimizing resources and logistic utilization. Computer vision can automate these tasks to reduce labour costs and improve efficiency by training deep learning models on appropriate data to perform knowledge-based tasks. Although fruit detection has been the focus of literature, yield prediction and harvest time estimation remain practical challenges for farmers. This is particularly important for high-value, highly perishable crops, such as raspberries, where contractual obligations require precise harvest timing. This paper addresses this gap by providing a learning-based framework for raspberry maturity detection and harvest time estimation. For this end, a phenology study of raspberry plants from three different cultivars was carried out and seven development stages were identified based on the BBCH-scale; for the first time, a phenology-based dataset of developing raspberry flowers and fruit was curated and made available publicly, which contains 1,853 high-resolution images and 6,907 manually labelled annotations. A comprehensive benchmark was developed from state-of-the-art object detection models, and finally, a tailored deep learning model capable of real-time inference was established that achieved 92.2% detection accuracy in the vertical farm field test. Depending on cultivar identity, the proposed model can estimate raspberry yield 28 to 33 days in advance of harvesting. 

This study addressed the critical challenge of precise harvest timing for highly perishable crops such as raspberries by developing an integrated phenology-based computer vision framework. To this end, a phenology study of three raspberry cultivars was carried out and seven phenology stages were identified based on the BBCH scale. Then, a phenology-based dataset of raspberries was curated and made publicly for the first time. This resource enabled the development and comprehensive benchmarking of deep learning models, from which a tailored YOLOv8-x model emerged as optimal, achieving a 92.2% detection accuracy in field validation and real-time operation. The dataset and models were also validated under realistic deployment conditions, with the model operating at frame rates compatible with real-time applications. 

The practical significance of this work lies in its ability to provide growers with a reliable tool for advancing yield forecasting by 28–33 days, a capability crucial for optimizing labour resources, fulfilling contractual obligations, and reducing postharvest waste. By demonstrating robust performance on consumer-grade hardware in an operational vertical farm environment, this research confirms the feasibility of deploying such deep learning computer vision solutions at scale. Given the global market value of raspberries and the increasing need for automation in horticulture, phenology-based detection frameworks such as those demonstrated here represent a timely and valuable contribution to the agri-food technology sector.

Should you have any questions, feel free to reach out to us at:  pjafary@torontomu.ca