What VineSense does
The system uses a camera mounted on a grape harvester to capture real-time images of the rows. Through segmentation neural networks and deep learning models, VineSense identifies the areas occupied by the clusters and estimates their volume. This process allows for continuous and non-invasive monitoring of production status.
How it works
- Image Acquisition: The camera installed on the machinery collects images of the clusters as it passes between the rows.
- AI Processing: The system has been trained on open-source datasets (WGISD) and images collected from the web.
- Segmentation and Markup: YOLOv4 and Unet models on EfficientNet-b0/b7 automatically identify and annotate the cluster areas.
- Real-time Operations: Optimized models (tiny_yolo3, PSPNet_ResNet34) guarantee fast performance on embedded devices such as nVidia Jetson TX2.
Technology
- Camera integrated into agricultural machinery.
- Neural segmentation for fruit area recognition.
- YOLOv4 + Unet on EfficientNet for automatic annotation.
- tiny_yolo3 + PSPNet_ResNet34 for real-time operations.
- nVidia Jetson TX2 as the embedded hardware platform.
Benefits
- Automation: Eliminates the need for manual monitoring.
- Precision: Accurate estimation of grape cluster volume.
- Efficiency: Continuous operation during harvesting activities.
- Scalability: Applicable to vineyards of any size.
VineSense is the ideal solution for wineries looking to optimize harvest management, reduce waste, and improve production quality.
