Sewing gesture image detection method based on improved SSD model

In this letter, the authors present a novel sewing gesture image detection method based on an improved single shot MultiBox detector (SSD) model. The deeper Resnet50 residual network replaces the VGG16 basic network of the original SSD model to improve the feature extraction ability. High and low level features are fused based on a feature pyramid network (FPN) for enhanced small‐target detection performance. The model is trained via transfer learning to resolve the small sample shortage problem. The proposed model shows an average precision of 88.69% on a sewing gesture data set constructed by the authors. The proposed model outperforms the Faster R‐CNN, YOLO, and SSD networks in terms of accuracy with acceptable operating speed on the same data set and fully satisfies the real‐time requirements for sewing gesture detection.