Comments (8)
from mctrans.
我使用的配置文件是mctrans_vgg32_d5.py
model = dict(
type='EncoderDecoder',
pretrained=None,
encoder=dict(
type="VGG",
in_channel=3,
init_channels=32,
num_blocks=2),
center=dict(
type="MCTrans",
d_model=128,
nhead=8,
d_ffn=512,
dropout=0.1,
act="relu",
n_levels=3,
n_points=4,
n_sa_layers=6),
decoder=dict(
type="UNetDecoder",
in_channels=[32, 64, 128, 128, 128],
),
seg_head=dict(
type="BasicSegHead",
in_channels=32,
num_classes=19, # 6
post_trans=[dict(type="Activations", softmax=True),
dict(type="AsDiscrete", argmax=True)],
losses=[dict(type="DiceLoss", softmax=True, to_onehot_y=True),
dict(type="CrossEntropyLoss"),
dict(type="FLoss")]),
aux_head=dict(
type="MCTransAuxHead",
d_model=128,
d_ffn=512,
act="relu",
num_classes=19, # 6
in_channles=[32, 64, 128, 128, 128],
losses=[dict(type="MCTransAuxLoss", sigmoid=True, loss_weight=0.1)]),
)
from mctrans.
from mctrans.
from mctrans.
hi,你可以提供你的数据在进入mctrans之前的shape吗?可能的话,提供下一张data还有mask,我这边debug一下
from mctrans.
你好,非常感谢你热心的解答,这个问题已经解决了,是维度的问题。但我自己的数据集里的GT中的像素有的值为255,是不需要计算指标的像素,这使我在使用损失时会遇到问题,只有交叉熵的损失有ignore_index的参数来让我忽略这些不需要计算的参数,Dice_Loss、FLoss、MCTransAuxLoss并没有这个参数,请问有什么办法可以解决这个问题么。
from mctrans.
Hi, 你那边可以对这几个loss进行更改,在计算loss的时候增加一个mask。或者你也可以在数据集里面简单对label进行remap, 例如将255 转为0,虽然简单,但是这会参与计算。
MCTrans/mctrans/data/datasets/base.py
Line 114 in 9b8b567
from mctrans.
非常感谢你的解答
from mctrans.
Related Issues (20)
- Where can I find the Appendix mentioned in the paper?
- The inference is none
- augment the training images with simple flipping HOT 2
- Dead 类指标为零 HOT 6
- CA module HOT 2
- The implement of NonLocalUNet is quiet different from the official implement HOT 3
- i had run this project and found an RuntimeError: shape '[32, 48, 16]' is invalid for input of size 22020096. HOT 8
- In the data you provide, the mask is pure black HOT 3
- 最后的指标和论文不一致 HOT 3
- 使用其他数据集出现的问题 AttributeError: 'ConfigDict' object has no attribute 'log_level'等等 HOT 1
- where is the final segmentation results graph?
- About the Auxillary Loss HOT 2
- 请问使用.npy格式的数据和使用.png格式的数据有区别吗 HOT 1
- DLL load failed HOT 1
- 'Single-channel predictions not supported' HOT 5
- proxy embedding
- Multi GPU Training
- Test pipeline
- 实验配置
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 mctrans.