We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Microplastic Classification in Holographic Images Using Swin Transformer V2
Summary
Researchers applied the Swin Transformer V2 deep learning model to classify microplastics in holographic images represented as HSL color-space composites of phase and amplitude information, achieving 91.65% accuracy and an F1-score of 91.79%. The study demonstrates that vision transformers can effectively leverage both spatial and color information from digital holograms for environmental microplastic classification.
Classifying microplastics with deep learning methods on integrated holographic images is more successful.By representing the microcosm along with its hologram phase and amplitude as an HSL color space, microplastic samples become rich in both spatial and color information. The Swin Transformer V2 which is famously known for its hierarchical utilization of self-attention and its success on minute and complex images, is chosen to perform exact classification. Results show the accuracy and F1-Score of 91.65% and 91.79%, respectively, thus reflecting a high level of true positive identification and good false positive avoidance. This study highlights the potential of vision transformers for environmental monitoring tasks and provides a framework for the automated microplastics detection via digital holography.