Software Development Kit
Gestoos gives you a powerful end-to-end platform to create rich behavior recognition and gesture interaction applications.Contact us
The Gestoos Developer Program provides you with powerful platforms, resources and training to build robust edge detection engines across multiple environments. From cars and cargo trailers to factories and ports, our technology solutions enable you to build and deploy best-in-class applications either entirely in-house or with our help. Leveraging Gestoos’ patented software, you can improve safety, increase operational efficiency and empower people to engage with their surroundings.
Product architecture component
From data set importation and data annotation to model training and evaluation, Gestoos CREATOR provides an end-to-end platform with no coding needs Detailed analysis for model validation is easily accomplished by linking results directly and visually with exact video frames. Build best-in-class models with a comprehensive easy to use tool leveraging patented cutting edge IP.
Our proprietary incremental learning with memory component maximizes accuracy thereby reducing training data needs. In addition, Gestoos CREATOR leverages semisupervised learning techniques that minimize the need for labeled data directly reducing annotation costs. And using non-annotated data has the advantage that additional specific task models can be defined and trained later leveraging incremental learning and a reduced supervised dataset.
Gestoos CREATOR has been designed by a team with many years of experience developing Computer Vision and AI based products. This extensive knowledge has been crafted into a complete toolkit that can be used by engineering teams with no specialized AI, Machine Learning or CV expertise. The whole model creation process is structured to help users to avoid common failures in data curation and selection, helping to balance categories and different types of metadata (subjects, view-points, scenarios).
The Gestoos CREATOR tool can be installed in the user company's cloud services account. In this manner data confidentiality and corporate IP is strictly maintained. This also allows the user to scale both the data storage and computing resources as needed.
Quick and agile development is achieved by leveraging pre-developed backbone and base models that dramatically reduce the overall end-to-end model creation time. Additionally, with incremental learning in each iteration only new data is processed reducing training time by orders of magnitude.
Our patented semi-supervised incremental learning methodology with internal memory component assures previous knowledge is not lost. It also generalizes better and far more accurately captures the long tail of low probability events. Gestoos' shape recognition technology delivers both higher accuracy and improved robustness.
Linux (Ubuntu 16.06 or higher)
CPU: Intel i5 2.7 ghz or equivalent(6th generation minimum required for RGB detection API)
8 GB RAM
Android System Requirement
Minimum CPU: Cortex A17 with Mali-T760 GPU
Recommended CPU: Cortex A73 with Mali G71 MP8
Supported from Android 7.0 up to 9.0
SDK API documentation
C++ software library
Touch free software
Hand Gesture Detection
Hand & Finger Tracking
Body and head interaction
Our SDKs are compatible with most of the depth cameras currently in the market. These are the models we currently support.
Intel RealSense depth camera D415
Intel RealSense depth camera D435
Occiptal Structure Core
Orbbec Astra Mini
Orbbec Astra Pro
Orbbec Astra Stereo
Orbbec Astra Embedded
Our pre-developed software solutions work with nearly all RGB and RGB IR cameras. Extreme FOV cameras (>150º) can also be used by adding incremental training data.
Laptops and PC Cameras
USB Web Camera