YOLOS Fashionpedia

valentinafevu/yolos-fashionpedia is an object detection model fine-tuned on the Fashionpedia dataset for detecting fashion items and accessories. It builds on the YOLOS architecture (You Only Look at One Sequence), a transformer-based approach that achieved state-of-the-art performance on COCO object detection.

The model was trained on Fashionpedia, a dataset of 46,781 images and 342,182 bounding-boxes covering 46 categories. The dataset includes 45,600 training images and 1,160 validation images, licensed under CC-BY-4.0. Its ontology comprises 27 main apparel categories (e.g., dress, jacket, pants), 19 apparel parts (e.g., collar, sleeve, pocket), and 294 fine-grained attributes. Research presented at ECCV 2020 demonstrated that instance segmentation models pre-trained on Fashionpedia achieve better transfer-learning performance on other fashion datasets than ImageNet pre-training.

The model detects categories such as shirt/blouse, top/t-shirt/sweatshirt, sweater, cardigan, jacket, vest, pants, shorts, skirt, coat, dress, jumpsuit, cape, glasses, hat, headband/head covering/hair accessory, tie, glove, watch, belt, leg warmer, tights/stockings, sock, shoe, bag/wallet, scarf, umbrella, hood, collar, lapel, epaulette, sleeve, pocket, neckline, buckle, zipper, applique, bead, bow, flower, fringe, ribbon, rivet, ruffle, sequin, and tassel.