Comments (3)
No. If directly running the code, what you have performed is the full-size attack (as shown in your figure), which corresponds to our main experiments (Table.1) in our paper. If you want to control the number of attack pixels, please modify the hyper-parameter 'temperature' (the default setting is -1, which means attacking all the pixels on the image). You can set whatever number you want as long as it doesn't exceed the limit. Also, to get better performance, I recommend smoothly decrease the temperature.
from adversarial-attack-on-person-reid-with-deep-mis-ranking.
I can not find the temperature value in your paper? Could you tell me the value of temperature you set in your paper? Or just tell me the suitable value.
from adversarial-attack-on-person-reid-with-deep-mis-ranking.
We use the symbol tau (Eq.2 in the paper) to represent temperature. And I will modify the helper of temperature to avoid the confusion.
If you only want to verify the results in the paper, you just follow the default hyper-parameters and ignore my above suggestion. So you don't necessarily worry about the value of temperature. The temperature is not an adaptive parameter and only relevant to the number of adversarial pixels. For example, you'll set the temperature as 8208 when you want to attack a quarter of the pixels in a given 288x144 image.
from adversarial-attack-on-person-reid-with-deep-mis-ranking.
Related Issues (17)
- There are some bugs in your code. HOT 1
- 对抗样本的可视化问题 HOT 4
- Black box attack HOT 1
- eval_freq?
- 运行时出错,我是直接下载的完整代码,但是在运行时出现如下错误: HOT 4
- Would D leak knowledge to G HOT 3
- Error while running with CUHK03 dataset
- The weight files HOT 1
- G model
- utils.py应该取消import cv2 HOT 1
- alignedreid.py 缩进没有调整好 HOT 1
- 论文中的mask问题 HOT 3
- Could ReID system be attacked by digital attack? HOT 3
- Feature Extraction HOT 1
- We are doing defense based on your code, but we don't know exactly how to customize our model HOT 2
- ???
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from adversarial-attack-on-person-reid-with-deep-mis-ranking.