Grapevine Inflorescence and Bunch Detection towards Automatic Yield Estimation using Computer Vision and Machine Learning

Dr. Muhammad Rizwan Khokher1, Dr. Dadong Wang1, Dr. Mark R. Thomas2, Dr. Everard J. Edwards2

1CSIRO Data61, Marsfield, Australia, 2CSIRO Agriculture and Food, Waite Campus, Australia


In viticulture, yield estimation is a key parameter, important for both vineyard management and many logistical aspects of winemaking and the wine industry. Yield can be estimated at different viticulture stages: inflorescence, pre-veraison, and harvest. At the inflorescence stage, where significant thinning of grape bunches does not occur, infield detection and counting of inflorescences following budburst have the potential to provide an early estimate of yield months before harvest. Whereas, at the pre-veraison and harvest stages, yield estimation can give more insight on how to efficiently allocate resources for winemaking.

This work presents an on-the-go approach (based on computer vision and deep learning techniques) for grapevine inflorescence and bunch detection in RGB images towards automatic yield estimation. To this end, datasets consisting of RGB videos with ground-truth bounding boxes for inflorescences and bunches present in the video frames were collected from multiple vineyards. Then a deep learning architecture was adapted to learn features from the images during training and detect the desired targets at the later inference stage.

The detection results are obtained for test images and videos which can be used to count the number of inflorescences and bunches, and ultimately for yield estimation. Based on the visual and quantitative results, we conclude that computer vision and machine learning based methods have the potential to provide yield estimation in viticulture.


Muhammad Rizwan Khokher is currently working as a ‘Research Engineer’ at Data61 CSIRO, Australia. He received a Ph.D. degree in 2018 from the University of Wollongong, Australia. He received an MSEE degree in 2012 from the National University of Science and Technology, Pakistan. He received a BSEE degree in 2009 from International Islamic University Islamabad, Pakistan. He was awarded full scholarships for his BSEE and Ph.D. studies. He has published 10 research papers in peer-reviewed journals and conferences, during his research career. His research interests include image/video processing, computer vision, machine/deep learning, and artificial intelligence.


AeRO is the industry association focused on eResearch in Australasia. We play a critical coordination role for our members, who are actively transforming research via Information Technology. Organisations join AeRO to advance their own capabilities and services, to collaborate and to network with peers. AeRO believes researchers and the sector significantly benefit from greater communication, coordination and sharing among the increasingly different and evolving service providers.