Comments (6)
I have some questions below:
- In your experiments, how many times did you split the trainval classes(760) to train(520) and val(240)?
- How do you evaluate your model on test classes? For example, is your one shot experiment result for a specific class calculated with each one of ten images as support image ?
So can you provide your split classes of train classes(520) and val classes(240)? With your classes split result and evaluation methods, a fair comparision can be conducted for people. @HKUSTCV
from fss-1000.
It is also confusing for me how the train set and validation set are used during training and the exact procedure to evaluate the model. From training.py
, there is no procedure to conduct model selection or hyperparameter searching based on the validation set. Perhaps it is in the evaluation
code which is not released. Could you provide the evaluation methods for a fair comparison?
from fss-1000.
I've added evaluation code in this PR: #13
from fss-1000.
@SMHendryx Have you evaluate the author published model (in readme.md) https://drive.google.com/open?id=1Vk0Pq8vOZrfrDtCISMcJmAQnt9jkXfPn in one-shot setting? I have tested on my own evalutation script, and also applied your PR #13 script by setting args.k=1, both experiments give the mean IoU of ~ 0.80, which is ~0.07 higher than the paper reported (0.73). Could you please kindly let me know your reproduction of the one-shot setting?
from fss-1000.
Could you publish the evaluate code? I think this repo is a good pipeline for few-shot segmentation. Thanks! @HKUSTCV
have you got that evaluation code?
from fss-1000.
@SMHendryx Have you evaluate the author published model (in readme.md) https://drive.google.com/open?id=1Vk0Pq8vOZrfrDtCISMcJmAQnt9jkXfPn in one-shot setting? I have tested on my own evalutation script, and also applied your PR #13 script by setting args.k=1, both experiments give the mean IoU of ~ 0.80, which is ~0.07 higher than the paper reported (0.73). Could you please kindly let me know your reproduction of the one-shot setting?
Would you please share that evaluation code with me ?
from fss-1000.
Related Issues (20)
- Reproducibility HOT 6
- The arrangement of the dataset-fewshot_data.zip???
- How to label dataset? HOT 1
- Problems in remote sensing imagery segmentation
- datadir in the training.py
- question about supplementary material. HOT 2
- 数据库错误 HOT 2
- What is your Directory Structure HOT 1
- What is the training shot
- Labels not normalized HOT 1
- Unseen classes
- instance annotation
- peregine_falcon/8.png is an RGB file, not a mask
- Wrong pixel size of some instances
- dataset mask error and image shape error
- config: docker HOT 1
- How to split dataset? HOT 2
- Some problems in the dataset HOT 3
- How to extend to C-way setting? HOT 1
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 fss-1000.