Austral Falcon provides a technological solution designed to digitize the processes that are currently carried out manually in the agricultural industry. Our platform applies state-of-the-art Machine Vision (MV), Deep Learning (DL) and Machine Learning (ML) technology to optimize decision-making based on quantitative information, acquired systematically and with high precision.
We provide a platform with the ability to organize all the relevant information on your crops and extract useful data that allows you to plan and make more informed decisions. This component is better described here.
Two Deep Learning models to detect two different variantes of grapes. These models were trained with rotation augmentations to guarantee its robustness.
This component is better described here.
The number of grapes counted in each cluster is adjusted by regression, which allows greater precision.
Deep Learning model with the ability to detect grape clusters in plants. The detection is performed using instance segmentation through a mask r-cnn neural network. It was implemented using the Meta Detectron2 framework and different data augmentation techniques were used to guarantee a robust model capable of offering good detection accuracy in different image or video capture conditions.
This component is better detailed here.
The cluster counter is a Deep Learning system that takes the detections of the cluster detector and feeds it into a tracking algorithm, thus allowing the clusters to be identified in a video and then counted.
The tracker used is a variant of DeepSort and the counting algorithm consists of a heuristic based on line crossings. More detail can be found here.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent