zjulearning/attentionzsl

Codes for Paper "Attribute Attention for Semantic Disambiguation in Zero-Shot Learning"

attentionzsl's Issues

The meaning of data of predicate-matrix-continuous.txt

Hi,

I am sorry to bother you about the meaning of data of predicate-matrix-continuous.txt.

How do you understand the " The number is the percentage of the time (between 0 and 100) that a human thinks that the attribute is present for a given class" which in the readme.txt of CUB dataset.

But In the awa dataset, data of predicate-matrix-continuous.txt had negative number，could you help me with the correct understanding of this?

Thank you very much!

I met some trouble on "configs/default.py"!

Hi, as you described, I downloaded the CUB dataset into "data/CUB/JPEGImages" and tried to run "python experiments/run_trainer.py --cfg ./configs/self_adaptation/VGG19_CUB_PS_C.yaml". However, I got "configs\default.py", line 72, SyntaxError: invalid syntax". Can you help me to fix it? Thanks!

The proposed_split file

Hello, sorry to bother you.
Could you please send me the proposed_split .txt file? I have already download the file from AwA2, but I didn't find the txt file in your code.
Thank you very much!!

How to get the " predicate-matrix-continuous.txt" of cub awa dataset?

Hi,

Thanks for your excellent work!
I have met the problem that how you get the " predicate-matrix-continuous.txt" of cub awa dataset,
if I want use imagenet dataset to train a model，could I get the file of imagenet?

Thank you again!

About the network arch you use for SUN and CUB dataset

About the results in the paper, I guess the networks you deploy for CUB and SUN (inductive setting) are not the VGG 19, but the GoogleNet or ResNet?

The poor perfermance of awa2 ps g ?

Hello, When I used "python experiments/run_trainer.py --cfg ./configs/self_adaptation/VGG19_AwA2_PS_G.yaml" and "python experiments/run_evaluator.py --cfg ./configs/self_adaptation/VGG19_AwA2_PS_G.yaml" to train and test on awa2 dataset, I got 0 accuracy?
Do you know why met the problem?

ckpt_epoch_15.pth
tr: 0.02500, ts: 0.00000, H: 0.00000

How to Check accuracy and loss after training

Hello!
I hope you are doing well. I have successfully train the network with CUB data set using ResNet101_CUB_PS_C configuration with self adaptation. After train the network when i run "experiments/run_evaluator.py" nothing is shown but the code execute successfully. How to check the Accuracy and other evaluation parameter in this code. Thanking you in anticipation.

The AWA2 dataset

Hello, Thanks for your excellent work!

You know the awa2 dataset that download from official website was so slowly, could you help me with
a baiduyunpan link?

Thank you very much!