Data Annotation is not only limited to image and video but now with recent advances in computer vision, it is extended to sensors even.
Many of us must be aware of image and video annotation techniques(if not you can go through these articles- on SuperAnnotate and LabelBox) such as bounding boxes, pixel-level accuracy, polygons, and many more such tools which have now enhanced the state-of-art. Sensor-fusion or Lidar annotation is another such technique. Lidar(Light Detection and Ranging) which measures distances using sensors between objects by illuminating them using a pulsed laser. This is used in many research applications by scientists.
Today we will be talking about Playment, a data annotator tool which allows all of this possible with workflow management.
What is Playment?
Playment is a complete data labelling platform generating training data for computer vision and machine learning models at scale build high-quality ground truth datasets. Playment was launched in November 2015 by Siddharth Mall, Ajinkya Malasane, Akshay Lal. It is headquartered at Bangalore, Karnataka, India.
Playment has a large community of trained annotators who have work distributed among them in the form of micro-tasks, formally known as microwork(this kind of work breaks down a large unified project into small tasks) that is done by people all over the internet. Annotators go through the platform and complete pending tasks and get points. The points can be exchanged in the form of vouchers on online e-commerce sites.
The platform is powered by a workforce of 300,000+ users which is managed by the human intelligence experts who build the tasks and deliver results with assured quality. It provides realtime ML-assisted pipelines with API support.
Features:
Image Annotation
Auto labelling facility among the 80 common classes present. The different tools available are 2D Bounding boxes, Semantic Segmentation, Cuboids, Polygons, Polylines, Landmarks.
Cuboids
Polylines
Polygons
3D Semantic Segmentation
Video Annotation
Provides 2D bounding boxes, Cuboids(image), Cuboids(LiDar), Polylines, landmarks
Sensor Fusion/Lidar Annotation
3D point cloud annotation
GT Studio
This dashboard allows us to set up and monitor customized workflows and build an end to end project management with playment workforce. ML engineers can use Python code to integrate their pipelines. Creation of Multiple user groups. Project progress and quality analytics with active feedback are also provided keyboard shortcuts, visualization tools and confusion metrics.
Use Cases:
- Deep learning assisted Semantic Instance Segmentation for Autonomous Vehicles
- Autonomous vehicles lane detection and driveable area.
- Full Pixel Segmentation
- Human Pose Estimation and Tracking
- Drone, CCTV and satellite surveillance
- Damage detection for car insurance
- Fashion, Gaming and Agriculture industry
Companies:
They empower high precision annotation services with complex customizations. Some of these companies and research institutions like Drive AI, Starsky Robotics, CYNGN, and UIUC, HELLA, Vayavision, INNOVIZ, Nuro (Building perception systems for robots), Daimler, Samsung, LG, Sony, Intel, Postmates, Siemens, Ouster(LiDar), AI AImotive, Alibaba, SAIC Motors, Continental.
