PyRetri/docs/MODEL_ZOO.md

Model Zoo

Here we provide the pre-trained models to help you reproduce our experimental results easily.

General image retrieval

pre-trained models

Training Set	Backbone	for Short	Download
ImageNet	VGG-16	I-VGG16	model
Places365	VGG-16	P-VGG16	model
ImageNet + Places365	VGG-16	H-VGG16	model
ImageNet	ResNet-50	I-Res50	model
Places365	ResNet-50	P-Res50	model
ImageNet + Places365	ResNet-50	H-Res50	model

performance

Dataset	Data Augmentation	Backbone	Pooling	Dimension Process	mAP
Oxford5k	ShorterResize + CenterCrop	H-VGG16	GAP	l2 +SVD(whiten) + l2	62.9
CUB-200	ShorterResize + CenterCrop	I-Res50	SCDA	l2 + PCA + l2	27.8
Indoor	DirectResize	P-Res50	CroW	l2 + PCA + l2	51.8
Caltech101	PadResize	I-Res50	GeM	l2 + PCA + l2	77.9

Choosing the implementations mentioned above as baselines and adding some tricks, we have:

Dataset	Implementations	mAP
Oxford5k	baseline + K-reciprocal	72.9
CUB-200	baseline + K-reciprocal	38.9
Indoor	baseline + DBA + QE	63.7
Caltech101	baseline + DBA + QE + K-reciprocal	86.1

Person re-identification

For person re-identification, we use the model provided by Person_reID_baseline and reproduce its resutls. In addition, we train a model on DukeMTMC-reID through the open source code for further experiments.

pre-trained models

Training Set	Backbone	for Short	Download
Market-1501	ResNet-50	M-Res50	model
DukeMTMC-reID	ResNet-50	D-Res50	model

performance

Dataset	Data Augmentation	Backbone	Pooling	Dimension Process	mAP	Recall@1
Market-1501	DirectResize + TwoFlip	M-Res50	GAP	l2	71.6	88.8
DukeMTMC-reID	DirectResize + TwoFlip	D-Res50	GAP	l2	62.5	80.4

Choosing the implementations mentioned above as baselines and adding some tricks, we have:

Dataset	Implementations	mAP	Recall@1
Market-1501	Baseline + l2 + PCA + l2 + K-reciprocal	84.8	90.4
DukeMTMC-reID	Baseline + l2 + PCA + l2 + K-reciprocal	78.3	84.2