Home

Awesome

DeepEmbeddingModel_ZSL

Tensorflow code for CVPR 2017 paper: Learning a Deep Embedding Model for Zero-Shot Learning

Li Zhang

Requirement

Python 2.7

Tensorflow > 1.0

Data

Download data from here and unzip it unzip data.zip.

Run

AwA_attribute.py will give you ZSL performance on AwA with attribute.

AwA_wordvector.py will give you ZSL performance on AwA with wordvector.

AwA_fusion.py will give you ZSL performance on AwA with attribute and wordvector fusion.

CUB_attribute.pywill give you ZSL performance on CUB with attribute.

GBU setting

ZSL and GZSL performance evaluated under GBU setting [1]: ResNet feature, GBU split, averaged per class accuracy.

AwA1_GBU.py will give you ZSL and GZSL performance on AwA1 with attribute under GBU setting [1].

AwA2_GBU.py will give you ZSL and GZSL performance on AwA2 with attribute under GBU setting [1].

CUB1_GBU.py will give you ZSL and GZSL performance on CUB with attribute under GBU setting [1].

aPY_GBU.py will give you ZSL and GZSL performance on aPY with attribute under GBU setting [1].

SUN_GBU.py will give you ZSL and GZSL performance on SUN with attribute under GBU setting [1].

ModelAwA1 T1usHCUB T1usH
DAP [2]44.10.088.70.040.01.767.93.3
CONSE [3]45.60.488.60.834.31.672.23.1
SSE [4]60.17.080.512.943.98.546.914.4
DEVISE [5]54.213.468.722.452.023.853.032.8
SJE [6]65.611.374.619.653.923.559.233.6
LATEM [7]55.17.371.713.349.315.257.324.0
ESZSL [8]58.26.675.612.153.912.663.821.0
ALE [9]59.916.876.127.554.923.762.834.4
SYNC [10]54.08.987.316.255.611.570.919.8
SAE [11]53.01.877.13.533.37.854.013.6
DEM (OURS)68.432.884.747.351.719.657.929.2
ModelAwA2 T1usHaPY T1usH
DAP [2]46.10.084.70.033.84.878.39.0
CONSE [3]44.50.590.61.026.90.091.20.0
SSE [4]61.08.182.514.834.00.278.90.4
DEVISE [5]59.717.174.727.839.84.976.99.2
SJE [6]61.98.073.914.432.93.755.76.9
LATEM [7]55.811.577.320.035.20.173.00.2
ESZSL [8]58.65.977.811.038.32.470.14.6
ALE [9]62.514.081.823.939.74.673.78.7
SYNC [10]46.610.090.518.023.97.466.313.3
SAE [11]54.11.182.22.28.30.480.90.9
DEM (OURS)67.130.586.445.1  35.011.175.119.4
ModelSUN T1usH
DAP [2]39.94.225.17.2
CONSE [3]38.86.839.911.6
SSE [4]51.52.136.44.0
DEVISE [5]56.516.927.420.9
SJE [6]53.714.730.519.8
LATEM [7]55.314.728.819.5
ESZSL [8]54.511.027.915.8
ALE [9]58.121.833.126.3
SYNC [10]56.37.943.313.4
SAE [11]        40.3  8.8    18.0  11.8 
DEM (OURS)61.920.534.325.6

PyTorch implementation

DeepEmbeddingModel_ZSL-Pytorch

Citing

If you use this code in your research, please use the following BibTeX entry.

@inproceedings{zhang2017learning,
  title={Learning a deep embedding model for zero-shot learning},
  author={Zhang, Li and Xiang, Tao and Gong, Shaogang},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2017}
}

References