Deep Neural Networks Estimate Weight and Height from Single Images
A new arXiv preprint explores automatic estimation of Body Mass Index (BMI), weight, and height from single images of people captured in uncontrolled environments. The researchers employ deep neural networks with single and multi-task learning, using RGB, depth-maps, pose-affinity maps, and edge-maps as input modalities. The study addresses challenges such as varying human pose, camera geometry, appearance, and backgrounds, using images from social networking websites. BMI is highlighted as a key health indicator linked to disease risk and longevity. The paper is available at arXiv:2607.26104.
Key facts
- arXiv:2607.26104
- BMI encodes weight and height characteristics
- Single image estimation in the wild is challenging
- Deep neural networks used with single and multi-task learning
- Modalities include RGB, depth-maps, pose-affinity maps, edge-maps
- Images sourced from social networking websites
- BMI helps predict disease risk and estimate longevity
- Paper explores automatic BMI, weight, and height prediction
Entities
Institutions
- arXiv