We are excited for the new release of the mobile image annotation tool
It has been more than a year since the first release of TaQadam image annotation app. It was originally built using an open-source tool at Oxford VGG Image Annotation tool. We are strong advocates for the open source technologies and intend ourselves to become one day a #mobilework community building inclusive societies.
2 million images annotated, 80+ active downloads and users, portfolio of AI training data projects with bounding boxes, polygons and semantic image segmentation
Reflecting on new clients’ requests, Annotators’ continuous feedback, we are happy to release TaQadam 2.0 on a Google Play today!
Why? For polygons, polyline and line annotation, the precision depends on a number of recorded points (x,y) and we mastered it by using touch annotation. We are increasingly focusing on semantic image segmentation and polygons, using format conversion for automatic calculation of bounding boxes, allowing companies to variate between classification, object detection and segmentation ML models with just one AI training.
Why? Training and Testing data for Computer Vision requires iterative cleaning, categorizing and preparing the data using a hierarchy of classes before finding a right fit for ML model. With the new feature we add several meta-tags to a bounding box or polygon, as well as classify the whole image for its ML fit using image-level tag system. Complex project with large number of classes, we are fit for a job!
Team Quality Control
AI-assisted quality control remains a dream if working on a unique project in robotics and industrial automation. We incorporated validation cycle for additional review. Gamification, new team roles with “extra” responsibility allowed us to increase performance, team’s understanding of class systems and ensure time management.
Incentives, incentives, incentives!
Transparency and scoring system decreases our turn around time! #mobilework is a foundation for financial inclusion.Data on past performance helped us understand how access to mobile earnings is a crucial for high retention rate and quality annotation. Our mobile app now incorporates the $ protocol. We are working on releasing API to further integrate with payment providers.
Reach out and share requirements for your next project with us
Originally published at: https://taqadam.io/taqadam-image-annotation-tool/