SeamFit: Towars Practical Smart Clothing for Automatic Exercies Logging (dl.acm.org)

SeamFit introduces a practical, garment-integrated approach to automatic exercise logging by embedding soft, stretch-sensitive sensors directly into clothing seams. Rather than relying on separate wrist or chest wearables, the system uses conductive threads and seam strain measurements to capture body movement patterns with minimal user burden. The authors pair this textile sensor hardware with a lightweight ML pipeline—time/frequency feature extraction and compact classifiers (suitable for on-device inference)—and include calibration/personnalization strategies to handle differences in body size, fit and movement style. The work is significant because it pushes smart clothing from lab demos toward real-world deployment: SeamFit emphasizes washability, comfort, and robustness across everyday use, and evaluates the approach on a range of common exercises in both controlled and in-the-wild settings. For the AI/ML community the paper highlights key technical trade-offs—sensor placement and sampling design, feature engineering vs. end-to-end models, and personalization to reduce labeled-data needs—while showing that seam-mounted textile sensing can achieve strong exercise-recognition and rep-counting performance without intrusive sensors. This opens paths for privacy-preserving, always-on activity logging for fitness, rehabilitation, and longitudinal behavior studies, while flagging remaining challenges in generalization, long-term drift and manufacturing scalability.