Closed-Loop of Data
Black Sesame provides closed-loop of data with effective management capability, empowering the quick algorithm iteration which could continuously improve the ADAS function by OTA
Edge Computing Module:
Shadow Mode Trigger
Based on the high computing power of our chips and decoupling of software and hardware, we are developing functions in mass production and shadow mode simultaneously on an integrated platform which unbounded the deployment of shadow mode from extra hardware and improved the capability of Trigger in defining scenes.
Data collection equipment:
Equipped with flexible data preprocessing system and data management platform, providing high bandwidth storage, flexible internal and external synchronization, supporting multiple cameras and a varity of image/video storage formats. A multi-sensor real-time recording system based on the powerful A1000 chip is convenient for customers to quickly build ground-truth acquisition vehicle systems for various usage scenarios, enhancing the efficiency, visualization and convenience of data processing.
Data Labeling:
Auto labeling, large model
A data labeling and management system deployed in a public cloud
involving ADAS all objectives labeling, surrounding parking BEV labeling,
semantics segmentation labeling in full scenes, Lidar 3D point cloud labeling and DMS labeling in a cockpit
Data Reinforcement:
Simulation
Emulating different unique scenarios for reinforcement of data under similar simulated scenarios and providing more corner case data for later NN training.
Multi-tasks Training:
Master network, feature network and single source of truth
Using data from full scenes and unique scenarios to re-train the model in order to improve the functionality and performance.
Shadow Mode:
Regression testing, gray release
Support shadow mode, empower to deploy regression testing and gray releasing strategy