Project Rana - Human Pose Keypoint Annotator (Thermal Imagery)

About the Role We are looking for annotators with computer vision or image labeling experience to perform 2D human pose (keypoint) annotation on thermal images within the client's Label Studio environment. You will identify people in each image and place a defined set of anatomical keypoints per person, including in cases of partial occlusion or ambiguous body positioning. Key Responsibilities Review each assigned thermal image and identify the number of people present. Create a separate annotation for each individual identified in the image. Position each keypoint as close as possible to the true center of the corresponding body part. Infer and place the expected anatomical position for keypoints that are obstructed or occluded. Indicate a confidence flag when keypoint placement is uncertain due to difficult body positioning or orientation. Submit completed annotations directly within the client's Label Studio instance. Participate in QA/rework cycles, revising annotations as needed based on reviewer feedback. Requirements Required: Prior experience with computer vision annotation, image labeling, or similar data annotation work. Solid understanding of human anatomy sufficient to accurately place and infer keypoints, even when partially visible. Sharp attention to detail and spatial reasoning, able to distinguish subtle differences in body positioning and orientation. Comfort working with thermal/infrared imagery (lower contrast, less visual clarity than standard RGB images). Ability to follow detailed annotation guidelines precisely and apply them consistently across large volumes of images. Good judgment for ambiguous cases (occlusion, unusual poses, overlapping people) and appropriate flagging of uncertainty rather than guessing silently. Familiarity with Label Studio, or willingness to quickly learn the platform. Reliable, self-directed work habits. Nice to Have: Experience with human pose estimation or keypoint annotation datasets (e.g., COCO keypoints format)