Skip to content

Latest commit

 

History

History
73 lines (50 loc) · 6.54 KB

MODEL_ZOO.md

File metadata and controls

73 lines (50 loc) · 6.54 KB
Raw

Model Zoo

More models with different backbones will be added to the model zoo.

Init models

Backbone Download
VGG-16 vgg16.pth

Attribute Prediction(Coarse)

Backbone Pooling Loss Top-5 Recall Top-5 Acc. Download (Google) Download (Baidu)
VGG-16 Global Pooling Cross-Entropy 13.70 99.81 model model, passwd: j9qd
VGG-16 Landmark Pooling Cross-Entropy 14.79 99.27 model -
ResNet-50 Global Pooling Cross-Entropy 23.52 99.29 model -
ResNet-50 Landmark Pooling Cross-Entropy 30.84 99.30 model -

Category and Attribute Prediction(Fine)

Backbone Pooling Loss Top-5 Cate. Recall Top-5 Attr. Recall. Download (Google) Download (Baidu)
VGG-16 Global Pooling Cross-Entropy 35.91 25.44 model -
VGG-16 Landmark Pooling Cross-Entropy 37.71 26.69 model -
ResNet-50 Global Pooling Cross-Entropy 42.87 29.37 model -
ResNet-50 Landmark Pooling Cross-Entropy 48.25 32.57 model -

In-Shop Clothes Retrieval

Backbone Pooling Loss Top-5 Acc. Download (Google) Download (Baidu)
VGG-16 Global Pooling Cross-Entropy 38.76 model model, passwd: wz69
VGG-16 Landmark Pooling Cross-Entropy 46.29 model -
ResNet-50 Global Pooling Cross-Entropy 41.81 model -
ResNet-50 Landmark Pooling Cross-Entropy 48.82 model -

Consumer-to-Shop Clothes Retrieval

Backbone Pooling Loss Top-5 Acc. Download (Google) Download (Baidu)
VGG-16 Landmark Pooling Cross-Entropy 7.18 model model, passwd: grfx

Fashion Landmark Detection

Backbone Loss Normalized Error % of Det. Landmarks Download (Google) Download (Baidu)
VGG-16 L2 Loss 0.0813 55.35 model model, passwd: 4ebx
ResNet-50 L2 Loss 0.0758 56.32 model

Fashion Compatibility Predictor

Backbone Dataset Embedding Projection Loss Fill-in-blank Acc Compatibility AUC Download (Google)
ResNet-18 Disjoint fully-connected layer Triplet loss, Type-specific loss, Similarity loss, VSE loss 50.4 0.80 model
ResNet-18 Disjoint learned metric Triplet loss, Type-specific loss, Similarity loss, VSE loss 55.6 0.84 model
ResNet-18 Nondisjoint fully-connected layer Triplet loss, Type-specific loss, Similarity loss, VSE loss 53.5 0.85 model

Fashion Segmentation

Backbone Model type Dataset bbox detection Average Precision segmentation Average Precision Download (Google)
Resnet50 Mask RCNN DeepFashion-In-shop 0.599 0.584 model

Fashion Virual Try-on

Model type Dataset Download (Google)
CP-VTON VTON GMM TOM