Genie Studio

Genie Studio is a one-stop development platform specially designed for embodied intelligent scenarios by Zhiyuan Robot. It has full-link product capabilities for data acquisition, model training, simulation evaluation, and model inference. It provides developers with standardized solutions from "procurement" to "training" to "test" and then to "push", which greatly reduces the development threshold and improves development efficiency. Through efficient data acquisition, flexible model training, accurate simulation evaluation and seamless model reasoning, the platform promotes the rapid development and application of embodied intelligent technologies. Genie Studio not only provides powerful tools, but also provides support for the large-scale implementation of embodied intelligence, accelerating the industry's leap to a new stage of standardization, platformization and mass production.

Demand population:

" Genie Studio is suitable for robot developers, research institutions, universities, and enterprises engaged in embodied intelligence research and application. It provides these users with a one-stop development platform that can help them efficiently perform data acquisition, model training, simulation evaluation and model inference, thereby accelerating the research and development and application of embodied intelligence technology. Through Genie Studio , users can get rid of data silos and computing power, allowing robots to continue to evolve in an environment of interweaving virtual and real, and promote embodied intelligence to transform from code executors to environmental cognizers."

Example of usage scenarios:

Robot developers use Genie Studio to quickly develop and deploy embodied intelligent algorithms to improve robot operation capabilities in complex environments

Research institutions conduct embodied intelligent research through Genie Studio , accelerating the transformation process from laboratory to practical applications

Universities use Genie Studio to carry out embodied intelligent teaching and scientific research projects to cultivate professional talents in related fields

Product Features:

Provide data solutions covering the entire life cycle of data, efficiently collect massive data, and a single machine with a single daily production capacity of up to 1,000 pieces, providing high-quality data support for model training

Provide self-developed and mainstream open source robot dock models, open up training, fine-tuning, quantization and deployment links, lower training thresholds, and support open source/private dataset collaborative training

Provide simulation evaluation capabilities, with 6000+ object assets and simulation scenarios, realizing user-side scene reconstruction, expert trajectory data acquisition, and visualization of evaluation results, helping to quickly verify algorithm performance and optimize models

Provides the ability to 'one-click real machine deployment', which can easily realize the seamless migration of algorithms from the cloud to the real machine environment, which is 2-3 times higher than traditional solutions inference performance, and accelerates the implementation of embodied intelligent scale

Support multi-ont and multi-end equipment management, and provide full-process functional services, including batch templated digital procurement task generation, full-link marking and visual audit, data set management, etc.

Equipped with a distributed architecture data pipeline, efficiently complete data cleaning, abnormal detection and other processing, supports dual modes of real-time processing and offline processing, ensuring massive data 'ready to be used'

It has VR and keyboard remote operation functions, supports quick verification of real machine remote operation function during simulation, accurate evaluation results, and the error between GO-1 model simulation test results and real machine results is less than 5%.

Integrate key capabilities such as model compiler, GDK real machine control system, low-code development system, application development framework, application release system, etc., to build an 'algorithm-deploy-management' closed loop

Tutorials for use:

Visit the official website Genie Studio to learn about platform features and usage guides

Register and log in to the platform, create projects and select appropriate robot models and datasets

Use the data acquisition module to collect or import multimodal data according to project requirements

Use the model training module to select pre-trained models or custom models for training and fine-tuning

In the simulation evaluation module, build simulation scenarios and perform algorithm performance evaluation and optimization

Through the model inference module, the trained model is deployed into a real machine environment for actual testing in one click.

Based on the evaluation results and actual test feedback, further optimize the models and algorithms to realize the iteration and improvement of embodied intelligent applications.