E-Commerce Product Image Segmentation
ImageBounding Box
Scope: total images/items, team size, timeline
Tasks: e.g., draw polygons around defects, label sentiment in text, transcribe audio
Size: e.g., “50,000 images” or “1M text tokens”
Quality measures: e.g., “double-blind review, 98% IOU threshold, weekly consensus meetings
For example
Labeled 15,000 drone images with bounding boxes and instance segmentation for a solar panel defect detection model. Worked in a team of 5, using Labelbox. Maintained >95% inter-annotator agreement via weekly calibration sessions and a detailed style guide. Delivered all batches within 2-week sprints
Scope: total images/items, team size, timeline Tasks: e.g., draw polygons around defects, label sentiment in text, transcribe audio Size: e.g., “50,000 images” or “1M text tokens” Quality measures: e.g., “double-blind review, 98% IOU threshold, weekly consensus meetings For example Labeled 15,000 drone images with bounding boxes and instance segmentation for a solar panel defect detection model. Worked in a team of 5, using Labelbox. Maintained >95% inter-annotator agreement via weekly calibration sessions and a detailed style guide. Delivered all batches within 2-week sprints
2024 - Present