A 3D Body from Eight Questions – No Photo, No GPU (clad.you)

A recent innovation allows the creation of a 3D body model using just eight questions, eliminating the need for photos or GPU-intensive processing. This small Multi-Layer Perceptron (MLP) operates effectively on a CPU, achieving height and mass accuracy of 0.3 cm and 0.3 kg, respectively. The approach addresses significant concerns around privacy, speed, and cost associated with traditional photo-based methods, while also improving upon previous models by accurately estimating body measurements based on user input data. This development is noteworthy for the AI/ML community as it not only provides a more efficient alternative to photo reconstruction but also enhances understanding of body shape variability beyond just height and weight. By incorporating additional parameters such as body build and shape, the model achieves a lower Mean Absolute Error (MAE) for body measurements compared to existing techniques. The integration of physics-aware loss in the training process refines mass calculations significantly, highlighting the model's capability to learn intricate relationships between parameters. Overall, this method represents a substantial advancement in generating realistic digital twins while reducing dependency on visual data inputs.