Are you passionate about using your Computer Vision Research Scientist expertise to make a positive impact in the world? If yes, this is the opportunity for you. We are seeking a highly skilled Computer Vision Research Scientist with experience in 3D computer vision scene / image understanding, 3D reconstruction and depth capturing sensors to join our team at Child Growth Monitor (CGM). Our mission is to prevent malnutrition in children by turning mobile phones into child-measuring devices. Through the use of augmented reality and artificial intelligence, we can accurately predict a child's height, weight, and middle upper-arm circumference to determine their nutritional status.
As a Computer Vision Research Scientist, you will be responsible for :
Image Data Preprocessing and Experimental Design :
- Image Processing : Perform preprocessing on RGB and depth data, including filtering, noise removal, normalization, and enhancement to improve feature extraction and model input quality.
- Depth Sensing and 3D Data Acquisition : Use Time-of-Flight (ToF) sensors (e.g., Huawei P30 Pro) and stereo depth sensors (e.g., Intel RealSense) to directly capture high-quality depth maps and 3D data of children. In some cases, utilize monocular depth estimation models to infer depth from a single RGB image when hardware depth is unavailable.
Camera Models and Calibration :
Sensor Calibration : Manage calibration of depth sensors (such as RealSense), including intrinsic and extrinsic parameter estimation, to ensure accurate 3D spatial measurements and sensor alignment.3D Modeling and Anthropometric Estimation :
3D Reconstruction from Depth + RGB : Generate accurate 3D models of children by combining depth maps and RGB images. Depending on the scenario, models may operate on 3D point clouds, depth maps only, or RGB images only, depending on availability and computational constraints.Height and MUAC Estimation : Estimate standing height and mid-upper arm circumference (MUAC) using either the 3D model, depth data, or RGB images, applying geometry-aware algorithms that maintain accuracy across various child poses, camera angles, and distances.Robust Model Development :
Variability Handling : Design algorithms robust to variations in lighting, background, body posture, camera viewpoint, distance, and motion artifacts.Outlier Detection : Implement methods to detect and correct depth map anomalies caused by occlusion, reflectivity, or sensor noise.Software Development and Real-Time Optimization :
Efficient Algorithm Design : Build lightweight, resource-efficient models for deployment on edge devices like smartphones and tablets.Real-Time Processing : Optimize inference pipelines for real-time or near-real-time execution using onboard compute.Data Collection, Annotation, and Augmentation :
Dataset Creation : Create diverse and representative datasets, covering various demographics, postures, lighting, and environments using RGB and depth sensors.GenAI for Data Augmentation : Leverage Generative AI techniques (e.g., GANs or diffusion models) to augment training data through synthetic data generation, including the creation of rare poses, occluded body parts, and challenging environmental conditions.Ethical Considerations : Ensure all data collection and processing comply with ethical standards and privacy regulations, especially when handling childrens data.Testing and Validation :
Cross-Validation : Use rigorous validation techniques including k-fold and hold-out validation to evaluate generalizability.Error Analysis : Perform detailed error analysis to identify sources of prediction error and guide improvements in model design.To be successful in this role, you should have :
Fundamental Skills in Computer Vision and Deep Learning :
Strong experience in 3D computer vision and deep learning, with 5+ years of research or industry experience.Proficient in Python, with solid experience in computer vision libraries such as OpenCV, and deep learning frameworks such as TensorFlow or PyTorch.Skilled in processing, cleansing, and verifying the integrity of depth maps and point cloud data for reliable analysis and modeling.Hands-on experience with Generative AI (e.g., GANs, diffusion models) for tasks such as synthetic data generation, data augmentation, and simulation of challenging scenarios to enhance model robustness.Experience in integration of an LLM-powered agent is a plus.Understanding of Human Anatomy and Biometrics :
[Nice to have] Anthropometry : Familiarity with human body proportions and measurement techniques to support accurate estimation of height, weight, and MUAC from image and depth data.Pose Estimation : Knowledge of 2D and 3D pose estimation algorithms to account for variations in posture and movement, especially in children.Experience with Sensor DataDepth Sensing : Experience in handling and analyzing data from depth sensors, including :
Time-of-Flight (ToF) sensors (e.g., Huawei P30 Pro)LiDAR-based devices (e.g., Microsoft Kinect)Stereo depth sensors (e.g., Intel RealSense D435i)Proficient in working directly with depth maps and point clouds, independent of the specific device.Collaboration and Interdisciplinary Understanding :
Experience collaborating with healthcare professionals (e.g., pediatricians, field workers) to validate measurements and improve the reliability of vision-based systems.Understanding user-centered design, especially in resource-constrained environments, considering usability for end users such as parents, caregivers, and healthcare workers.Research and Problem-Solving Skills :
Competence in conducting literature reviews, synthesizing academic and applied research to inform model design.Skilled in experimental design for validating algorithms and measurement systems.Proven innovative thinking with the ability to design creative and practical solutions to complex technical challenges.Strong communication skills, both written and verbal, for collaboration, reporting, and presentations.Degree and Professional Experience :
A master's or PhD degree in Computer Science, Electrical Engineering, Artificial Intelligence, Computer Graphics, or a related field, and / or equivalent professional experience in relevant roles.[Nice to have] Prior experience in building and deploying Android AR / VR applications.Work Logistics and Collaboration :
Willingness to travel to data collection sites in India or other international locations for field validation and deployment activities, typically once or twice a year.Must have the flexibility to align working hours with the German time zone once or twice a week to enable effective collaboration with international team members.Why should you join us ?
This is a consultant position. The duration of the contract is one year, and the position is fully remote, available to candidates located in time zones from GMT+2 to GMT+7. The working language for this role is English. We encourage you to apply even if you dont meet all the requirements mentioned above.
As part of the Welthungerhilfe team, you will have the opportunity to work with a global organization committed to ending hunger and malnutrition by 2030. Most importantly, you will have the satisfaction of knowing that your work is helping to save the lives of millions of children.
More than this competitive remuneration, youll be doing the work of a lifetime with Child Growth Monitor. Its a moonshot project just like autonomous car driving that no one has figured out completely yet. Your contribution can save the lives of 3 million children under age 5 who die from untimely diagnosis of malnutrition. Apart from this, youll have the opportunity to collaborate and publish papers with WHO, UNICEF, Microsoft, Tilburg University, respected independent researchers, etc.. You enjoy Flat culture, a friendly environment, teams making decisions, and a work-life balance.
(ref : hirist.tech)