aiData Workflow
With the aiData Recorder, ensure that any multimodal sensor setup is calibrated and synchronized properly during data collection.
Oversee your data collection campaign to monitor potential errors and check sensor diagnostics real-time. Flag events automatically, if your ADAS stack is integrated with the pipeline.
|
|
All data collected is automatically uploaded to the aiData Versioning System, where developers can replay and analyze recordings, and send interesting sessions for automatic annotation. |
The aiData Auto Annotator creates high-precision multi-modal annotations for static objects and dynamic traffic participants. The aiData Annotator achieves 90%+ overall precision. |
|
With the aiData Versioning System, curate diverse datasets for a variety of use-cases, enjoying text- and image-based searchability via an intuitive Ul. |
|
|
Understand the strengths and weaknesses of your data with aiData Metrics, and create illustrative reports of your training and testing data campaigns. |
Benefits of integrating the aiData pipeline
On the team level
-
Track and manage the data journey at one place to reduce complexity
-
Speed up development with high-level of automation in the data toolchain
-
Achieve high-precision annotation, within hours of the recording
-
Isolate mistakes and understand the strengths of your data, even during real-time recording
-
Complete transparency and traceability over the entire data flow
-
Curate datasets with latest AI-based technologies, as text, scenario and video-based searching
On the company level
-
Save costs by relying on an automated workflow, including automated annotation and AI-based features for data curation
-
Deploy aiData on-premise for the highest security or in the cloud for easy collaboration between global teams - while enjoying full ownership over your data
-
The aiData team is ready to adapt and customize the aiData pipeline to any needs, such as developing new features and integrating it to existing tools
-
Focus on teams with the highest added value, no need for manual annotators
-
Flexible business model from full service to software licensing
Selected use-cases
Query and manage real-word data
Manage data recorded by external loggers or the aiData Recorder, all utilizing the auto tagging of the aiData Versioning System. With our system, explore the details of the whole journey such as time-of-day, domain and special environment, and search for interesting events, based on object class, traffic sign types, traffic light states, among many others. By this, curating diverse datasets for a variety of use-cases can be achieved smoothly and efficiently.
Collect your data Synchronized and well-calibrated
Record data with the aiData Recorder with built in tools to ensure synchronization and calibration between sensor streams. It also enables real-time sensor diagnostics and sensor calibration to ensure that potential errors can be identified and corrected during anytime in the data collection campaign.
Auto annotate real-world data
Annotate data automatically with the aiData Auto Annotator, including dynamic traffic participants and static objects, utilizing AI algorithms and a GPU cluster. Collect and annotate continuously, while receiving real-time feedback on the accumulated dataset for efficient adjustments.
Create your own metrics
Measure development progress against requirements and dive into real-time insights and data gap analysis. aiData Metrics support environment detection benchmarks, such as object tracking and confusion matrix, and vehicle function benchmarks, such as Autonomous Emergency Braking, to monitor the performance of your algorithms.
Generate synthetic data to address data gaps
Create great volumes of diverse synthetic data, with the variability needed to replicate real-world conditions for training and testing your AD software. Generate synthetic sensor data with plausible scenarios and event sequences, and vary environmental and road conditions.
Create high-precision maps from raw sensor data
Generate customized HD map providing a solid basis from entry level ADAS to complex automated driving systems without the need for third-party map source. Map complex roads, with lane level precision and enrich them with traffic sign and light information to prepare vehicles for the upcoming scenarios.
Create 3D environments with neural reconstruction
Create high-fidelity reconstruction of 3D assets in a highly scalable way with neural rendering, all without relying on manual labor.
