Feature Point Extraction of Garments based on Deep Learning
Choosing informative, discriminating and independent features is an important step for effective algorithms in pattern recognition and image classification. The goal of this investigation is to advance the state-of-the-art of articulated training previous excited methods and implement our dataset and modify. Evaluation is done on 2D images classified in 5 categories and pose of apparel and visibility is different. To encourage future studies, we introduce a key point detection dataset provided by Alibaba e-commerce with over 45K images, where each image is labeled with 24 landmarks. The proposed approach performs same as best known multi-person pose estimation results while demonstrating key point performance on the detection of garment.
Keywords - Feature extraction, Deep Learning, Convolutional Neural Network, ResNet, Tensorflow