Some of our vertical Use Cases
Autonomous Vehicle
Case Study: Autonomous Vehicle Startup
Hot startup in Autonomous Vehicle is building cutting edge perception systems for OEMs. The needed high-quality data annotation workforce to label videos for semantic segmentation, bounding box, Lidar 3D point cloud for segmentation and object tracking with sparse point cloud data.
Our Solution
We iterated on the pilot dataset to come to a high-quality instructions set and provided guidance about task splits to optimize workforce productivity. Our Solutions architects helped operate the labeling platform in the customer's account and our full stack developer provided minor enhancements to the annotation UI for the quality assurance pipeline to capture feedback. We provided quality labels for 2D instance segmentation, 3D Point cuboids and segmentation.
Manufacturing
Case Study: Chemical Multinational
Chemical industrial giant was bringing AI to the edge with the ability to remotely manage predictive events. They started using computer vision to monitor processes and identify defects early in the process life cycle. The annotation needed expertise in chemical processes.
Our Solution
Objectways process engineering expert worked with the customer to optimize labeling pipeline and train other team members with a series of pilot tasks. The task involved complex bounding box and semantic segmentation tasks pi. with rigorous audit controls. The project was delivered in 2 months with 5oK+ labels.
Drone Imagery
Case Study: Drone Imagery Start up
A leading drone imagery customer who uses aerial imagery analyzed via computer vision to help their customers in insurance, solar, utilities and government sectors. They continue to amass new aerial imagery and use transfer learning techniques to improve machine learning models to save time and money. They needed an object detection annotation pipeline for roof measurement.
Our Solution
The computer vision team presented multiple tool options for semantic segmentation to ensure all necessary labels are captured along with respective attributes. Objectways solutions architect also helped set up an annotation data lake to ensure no redundancy in data labeling and resulted in lowering the cost. The dedicated CV annotation team supports this initiative on an on-demand basis.
Financial_Services
Case Study: Leading Fintech bank
In the investment sphere, applying tags to highlight the main topics covered by text, or topic modeling, is valuable when analyzing earnings calls to establish a main theme, or to compare against previous, similar calls to identify trends. The customer wanted to tag large corpus of data to build a graph database.
Our Solution
Objectways built a custom workflow to suit the Customer's needs and trained the workers on the pilot workload to ensure high-quality instructions. We completed 150K documents tags successfully in 8 weeks.
News & Media
Case Study:Video Archive Transcription
Thomson Reuters (US) has a large collection of news archive videos collected over the last 5o+ years. Reuter has developed machine learning models to detect speaker, language and transcribe videos into text. Reuter wanted to collect large quantities of ground truth data to improve machine learning models and for model evaluation.They were looking for a scalable labeling solution.
Our Solution
Thomson Reuters (US) uses AWS cloud for their infrastructure. They looked at multiple labeling providers including Amazon SageMaker Ground Truth but did not find a suitable annotation UI to complete the task for speaker and language identification. Objectways was able to develop a user friendly annotation UI hosted as Amazon SageMaker Ground Truth template for 3rd party labeling companies. The UI allowed annotators to watch archived video as a task and record certain portions of clips with appropriate language or speaker. The labeling project was successfully completed within 3 months and enabled Reuters to use collected labels to improve their Machine Learning models.
Healthcare and Life Sciences
Case Study: Electronic Health records
Electronic Health records include Doctor's notes, patient history, prescriptions. Unlocking Medical, PII and Disease related terms within this unstructured data can help understand broader patterns for analytics for patient intervention. Healthcare start up was on a mission to build Machine Learning models to create insights from unstructured EHR. They needed large amount of labeled data to train ML Models.
Our Solution
Our NLP team has a seasoned Medical experts including nurses and consulting doctors who were able to provide 98% quality metrics(precision/recall/F1/Keppa score).