Comments (7)
For the first question, we use CME to imporve MPSR's classfication, so we use it in refinement branch, not another branch.
For the second question, we view margin as inter-class distance as margin. Hope this paper can help u. (https://arxiv.org/pdf/2005.13826.pdf)
Sorry, I still don't understand. CME is meta-learning based method, however, MPSR is fine-tuning method without support branch, so how to implement Feature Disturbance while there is no support mask?
from cme.
Because of limited time, we only train our model on COCO dataset with MPSR. Hope this can help you.
from cme.
Sir, thanks for you quick reply. Do you mean you apply the method on top of MPSR, but actually I am puzzled how you apply CME to two stage detectors.
MPSR is based on FPN which only take one input, but CME needs support branch and query branch.
So you equip FPN with an extra support branch and train it with the meta learning pipeline right? Could you tell me the baseline performance of that.
Also I am puzzled about the
Argmax L_mrg is equivalent to increase intra and decrease inter, but this will lead the margin decrease? Why you say it will increase margin? Could you explain to me. Thanks in advance!
from cme.
For the first question, we use CME to imporve MPSR's classfication, so we use it in refinement branch, not another branch.
For the second question, we view margin as inter-class distance as margin. Hope this paper can help u. (https://arxiv.org/pdf/2005.13826.pdf)
from cme.
For the first question, we use CME to imporve MPSR's classfication, so we use it in refinement branch, not another branch.
For the second question, we view margin as inter-class distance as margin. Hope this paper can help u. (https://arxiv.org/pdf/2005.13826.pdf)Sorry, I still don't understand. CME is meta-learning based method, however, MPSR is fine-tuning method without support branch, so how to implement Feature Disturbance while there is no support mask?
Sorry, we use Feature Disturbance to disturb feature. Because mask reflects object position, we use it. You can disturb corresponding object in image or corresponding feature. Because of limited time, we don't more experiment in it. We'll try to do some research in it.
from cme.
@Bohao-Lee 你好 针对Retina的第二个问题, 我也有疑惑,我想问下为什么margin loss的公式不是 argmin Lmrg? 因为您的论文底下也要提到要 增大类间距离D_inter 减少同类之间距离D_intra 包括你的loss 公式也是 argmin L = Ldet + Lmrg, 这不是和argmax L_mrg矛盾吗? 希望您能赐教。
from cme.
@Bohao-Lee 你好 针对Retina的第二个问题, 我也有疑惑,我想问下为什么margin loss的公式不是 argmin Lmrg? 因为您的论文底下也要提到要 增大类间距离D_inter 减少同类之间距离D_intra 包括你的loss 公式也是 argmin L = Ldet + Lmrg, 这不是和argmax L_mrg矛盾吗? 希望您能赐教。
十分感谢您对我们工作的关注,论文中由于疏忽导致公式错误,这里应该修改为arg min Lmrg,具体实现可以参考代码,对您造成的困惑还请您谅解,十分感谢您的指正。 @Wei-i
from cme.
Related Issues (20)
- Questions about model size HOT 1
- Questions about nGPU
- Hi Can you provide the t-SNE code? HOT 2
- 能否共享一下t-SNE的代码?并分享一下如何使用。感谢 HOT 2
- Some question about MPSR baseline HOT 4
- About the process datasets
- How many GPUs will be used for training the code? HOT 9
- Feature disturbance applied on both base&novel? HOT 1
- Question about .weights HOT 2
- About the category and confidence after the test
- 前辈 检测结果图中显示类别求指教!
- How long is the model trained by using two Nvidia Tesla V100 GPUs as mentioned in the paper and the github issue?
- occur the NAN when training the Net using the pascal voc HOT 2
- How to change the framework from Yolo to Faster R-CNN ? HOT 1
- Where is the code for 'Feature disturbance'
- RuntimeError: shape ‘[30, 5, 6, 13, 13]’ is invalid for input of size 5070 HOT 3
- ModuleNotFoundError: No module named 'core'
- RuntimeError: shape '[64, 15, 845]' is invalid for input of size 1622400 HOT 1
- As the training time increases, Proposals=0, is this normal?
- Problem:Get result If you want to get the result of model, run:
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 cme.