BrightAnnotate an Automated Data Annotation Engine by BrightDrive
What IS bRIGHTANNOTATE BY BRIGHTDRIVE?
Enhance and accelarate the deployment of your annotation algorithms
BrightAnnotate® is an automated data annotation AI engine that provides pre-labeled annotations for scenes coming from different regions based on different kinds of sensors. BrightAnnotate is usually combined with a manual annotation pipeline to provide 100% ground truth annotations.
Using deep learning models, BrightAnnotate’s four software modules cover traffic participants, road, traffic control information and static environments. It processes data from lidars, 360° cameras and GNSS to ensure accurate annotations for advanced perception systems.
Drag and Drop and deploy your semi-automated annotation algorithms in seconds
brightdrive gallery
brightannotate in details
BrightAnnotate software modules for semi-automated data annotation
BrightAnnotate® accelerates the development, testing and validation of advanced assistance systems. The software modules can operates independently of the sensor set up of your vehicle.
BrightAnnotate® modules:
Dynamic Envinronment Fusion (DEF)
This module generates dynamic traffic objects in the scene:
- Covers 6 classes: cars, pedestrians, trucks, buses, cyclists, and motor bikes.
- Avg. Model mAP on the trained ODD: 80%
- Generates output in JSON file format that contains:
1. 3D Boundary Box points in ego coordinates.
2. Velocity.
3. Acceleration.
4. Object tracking id.
Road Mode Generation (RMG)
This module generates road objects like road boundaries, lanes, and pavement markings.
- Covers 25 classes between lane lines type and color, road boundaries, and pavement markings.
- Avg. mAP of lanes on the trained ODD: 85 %
- Avg. mAP of boundaries on the trained ODD: 90 %
- Avg. mAP of pavement marking on the trained ODD: 80%
- Generates output in JSON file format that contains:
1. Road topology in 3D.
2. Road boundary 3D points with reference to ego vehicle position.
3. Road boundary type.
4. Lanes lines 3D points with reference to ego vehicle position.
5. Lane line type.
6. Lane line color.
7. Pavement marking 3D center point and association to lanes.
8. Pavement marking type.
Static Environment Fusion (SEF)
This module generates static objects in the scene:
- Covers 8 classes between simple and extended objects:
1. Construction: barrier, bridge and tunnel.
2. Object: pole, utility pole, traffic cone, and pothole.
3. Nature: vegetation.
- Generates output in Json file format that contains:
1. Static object type: static/ extended.
2. Static object class.
3. Static object 3D location with reference to ego vehicle position.
4. Static object 3D projection points with reference to ego vehicle position.
5. Object tracking id.
- Avg. Model mAP on the trained ODD : 82 %
Traffic Information for Signs & Light (TIS/L)
This module generates a wide range of traffic signs and traffic lights objects.
- For lights
- Covers 15 classes:
- States:
1. Red
2. Off
3. Yellow
4. Red yellow
5. Green
- Generates output in JSON file format that contains:
1. Traffic 3D location with reference to ego vehicle position.
2. Traffic object tracking id.
3. Traffic light state.
4. Traffic light shape.
- Avg. mAP of states on the trained ODD: 80 %
- Avg. mAP of shapes on the trained ODD: 75 %
- For signs
- Covers 186 classes of Europe traffic signs classes between 166 main types and 20 supplementary types.
- Generates output in JSON file format that contains:
1. Traffic sign 3D location with reference to ego vehicle position.
2. Traffic object tracking id.
3. Traffic sign type.
- Avg. mAP on the trained ODD : 92 %
- For more information about the classes, and attributes of the annotation and the ODD of the datasets, Please feel free to contact us.
- Disclaimer: The performance metrics for the neural networks were estimated based on the datasets collected by BrightDrive in similar ODD conditions to the ones that were used to train the network. Metrics will differ based on the quality and conditions of any newer datasets used to infer the model. In general, due to our backpropagation pipeline, we are continuously enhancing the performance of our models with newer data to cover as much ODDs as possible
ABOUT brightdrive
BrightDrive is a company owned by Brightskies based in UAE, and is focused on development of autonomous driving systems. BrightDrive develops Driver Assistance and Fully automated driving solutions for different forms of vehicles. BrightDrive also provides fully autonomous Mobility as a Service solutions for logistic chains and smart cities.
The mission of BrightDrive is to revolutionize mobility by developing cutting-edge solutions that prioritize safety, efficiency, and reliability. From advanced perception algorithms to precision localization techniques, our software and hardware platforms are engineered to enable safe and efficient autonomous operation in diverse environments.
Product lines:
FOLLOW Brightdrive
Brightskies Headquarters
Palm City Towers
Talateen El-Gedid St:
Mandarah Qebly
Alexandria, Egypt
[Reach the address on Google Maps]
www.brightskiesinc.com
For more information, contact dSPACE and Intempora
Please, fill the form below.
