Hi all,
We will have a student presentation upcoming Wednesday (13.03.2024) at 1:00 pm.
Laurent Hug, Master student in the CIX group, will be presenting his final Master thesis talk. You can find the thesis title and abstract below.
The presentation will take place in E1.7, room 001.
Best regards,
Marie
------------------------------------------------------
Title:
StitchSense: A Deep Learning Approach to Counting Knitting Stitches in
Real-Time using Hand-Based IMU Sensors
Abstract:
While knitting enjoys wide popularity as a traditional craft, it has received limited attention in the realm of automated pattern recognition or technical analysis. This thesis examines the unique application of machine learning techniques for real-time knitting stitch counting. We create and introduce a comprehensive hand-labeled knitting motion dataset as a foundation for our analyses. Our exploration, which encompasses both classic and advanced neural network architectures, demonstrates a notable advantage of individualized models (achieving an average within-user F1-score of 0.71). However, cross-user generalization proved challenging (with an average F1-score of 0.37), a finding that prompts future research into adaptable feature extraction or fine-tuning approaches. With potential applications in digitized
pattern matching and knitter technique analysis, the dataset developed within this thesis expands opportunities within this emerging domain. We further demonstrate the feasibility of this novel concept through the development of a robust pipeline for data collection, feature engineering, model training, and real-time counting. Ultimately, this work highlights the potential of machine learning to augment traditional crafts, setting the stage for future studies focused on innovative assistive technologies.