lsaiml / cavinet

Code for Deep Cross modal learning for Caricature Verification and Identification (CaVINet), ACM MM, 2018

CaVINet

Code for Deep Cross modal learning for Caricature Verification and Identification (CaVINet), ACM MM, 2018

Dataset

Please check CaVINet_Dataset folder for the visual as well as the caricature images (both present in independent folders). Note : The CaVI Dataset must be used for research only and not for any profit/commercial purpose. Dataset Directory structure:

CaVINet_Dataset
└───Real
│    └───Angela_Merkel
│    │   │   Angela_Merkel_r_0.jpg
│    │   │   Angela_Merkel_r_1.jpg
│    │   │   ...
│    │   
│    └───Bill_Gates
│    |   │   Bill_Gates_r_0.jpg
│    |   │   Bill_Gates_r_1.jpg
│    |   |   ...
│    |
│    └─── ...
│
└───Caricature
    └───Angela_Merkel
    │   │   Angela_Merkel_c_0.jpg
    │   │   Angela_Merkel_c_1.jpg
    │   │   ...
    │   
    └───Bill_Gates
    |   │   Bill_Gates_c_0.jpg
    |   │   Bill_Gates_c_1.jpg
    |   |   ...
    |
    └─── ...

Requirements

• Keras with TensorFlow backend
• Keras-VggFace - pip install keras_vggface
• python 2.7

Architecture

Pre-trained weights

Link to be added soon. Please stay tuned.

cavinet's Issues

Lacking rcmalli_vggface_tf_vgg16.h5

On Line 214 of train.py:
model.load_weights( 'untied_softmax_weighted_error_33_33_33_lagmul_0.2_aug.h5')
I can not find this file. Could you please upload it?

where is the data_instances2 folder?

in the train.py , the trainging parameters includes files in the folder of data_instances.
But i can not find the folder.
Can you help me ?

untied_solfmax_weighted_error_33_33_33_lagmul_0.2_aug.h5 required

I can't find the pre-trained weights untied_solfmax_weighted_error_33_33_33_lagmul_0.2_aug.h5 in program. Could you please help with this?

.txt and .npy file required

I did not find any .txt and .npy file in your dataset that has been used in program. Can you please help me with this?

Dataset

Can you please provide the data set?

one question about train.py

I really appreciate your work. When I read the code, I found a few puzzles.
Firstly, what do those 2 lines of code mean? (About 70 lines in train.py)
train_labels[count, 0] = (np.where(identities == i[0]))[0][0]
train_labels[count, 1] = (np.where(identities == i[2]))[0][0]
Secondly, I can't find this file untied_softmax_weighted_error_33_33_33_lagmul_0.2_aug.h5(about 215 lines in train.py)

identities_195.npy & /home/btp17-18-2/Data/ required

I did not find identities_195.npy and /home/btp17-18-2/Data/ file in program. Can you please help me with this?

Can you please upload the requirements.txt file to run this code

@

can you offering the model file?

dataset release?

I greatly appreciate your work.
can you release dataset?