Cloth Classification for the company slow fashion. All data and model trained on it belong to slow fashion and cannot be published.
For quick testing
kaggle Fashion MNIST
Bigger dataset (7gigs)
kaggle clothing dataset
Biggest dataset (250gigs)
DeepFashion2, Form to get the code
Deep Learning for Clothing Style Recognition Using YOLOv5 This paper do Clothing Style Recognition with YOLOv5 and R-CNN, the metric used are average precison, mean average precison, recall, F1-score, model size, and frame per second. They Evaluate different architecture and want to be deployable on mobile device.
DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations
Deep Residual Learning for Image Recognition Best paper about residual neural network, should be better than VGG-16 but a lot deeper
DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations presents the DeepFashion dataset, a large-scale dataset for benchmarking tasks in fashion recognition and retrieval. Measure:
-
Category and Attribute Prediction: Top-k accuracy (e.g., Top-1, Top-5): Measures whether the ground truth label is among the top-k predicted labels. Mean Average Precision (mAP): Especially for attribute prediction, to account for multi-label outputs.
-
Landmark Detection: Normalized Error (NE): Distance between predicted and true landmark locations, normalized by image size. Detection Rate @ ε: Percentage of correctly predicted landmarks under a specific error threshold ε.
-
Clothes Retrieval: Top-k retrieval accuracy: Measures whether the correct match is among the top-k retrieved items. Recall@k: Proportion of queries for which the correct item appears in the top-k results.