ZSL_PP

Mohamed Elhoseiny*, Yizhe Zhu*, Han Zhang, Ahmed Elgammal, Link the head to the "peak'': Zero Shot Learning from Noisy Text descriptions at Part Precision, CVPR, 2017

This code is implemented by Yizhe Zhu and Mohamed Elhoseiny.

Processed feature Data:

You can download the dataset CUB2011 and NABird.

Raw wikipedia article data:

Raw wikipedia article data of CUBird and NABird, as well as detailed merging information of NABird, can be obtained here.

Trianed Models:

Trained models reproduce the results in the paper.  

Testing, reproducing the results in the paper

---

ZSL_Test(Dataset = 'CUBird' or 'NABird', splitmode = 'Easy' or 'Hard', ImgFtSource = 'DET' or 'ATN')

splitmode = Easy or Hard splits defined in Section 4.1 in the paper

CUNBirds Easy split in Table1

---

ZSL_Test('CUBird', 'Easy', 'ATN') ### ATN means using groundtruth part annotation
Dataset: CUB2011 Easy ATN
Model: trained_models/CUBird_Easy_ATN.mat
Load Testing set
test_acc = 43.5049%

---

ZSL_Test('CUBird', 'Easy', 'DET') ### DET means using the detected parts instead of GT parts.
Dataset: CUB2011 Easy DET
Model: trained_models/CUBird_Easy_DET.mat
Load Testing set
test_acc = 37.5725%

NABirds Easy/Hard split in Table3

---

ZSL_Test('NABird', 'Easy') ### Easy means category-share splitting
Dataset: NABird Easy DET
Model: trained_models/NABird_Easy_DET.mat
Load Testing set
test_acc = 30.5937%

---

ZSL_Test('NABird', 'Hard') ### Hard means category-share splitting
Dataset: NABird Hard DET
Model: trained_models/NABird_Hard_DET.mat
Load Testing set
test_acc = 8.1349%

Training

ZSL_Train(Dateset, Splitmode, ImgFtSource, lambda1, lambda2, GPU_mode)
is the command to train the model using a particular setting.
% For example ZSL_Train('CUBird', 'Easy', 'DET', 100000, 10000, true), trains on the CUBirds dataset on the Easy split and using the detected part boxes. , lambda1=100000, and lambda2=10000, and GPU_mode=true (using GPU mode for training). If false, the training is done on CPU.