Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
lsyzz ae2272bcb3 | 10 months ago | |
---|---|---|
.. | ||
_base_ | 10 months ago | |
ann | 1 year ago | |
attention_unet | 3 years ago | |
bisenet | 1 year ago | |
bisenetv1 | 1 year ago | |
cae | 1 year ago | |
ccnet | 1 year ago | |
danet | 1 year ago | |
ddrnet | 1 year ago | |
decoupled_segnet | 1 year ago | |
deeplabv3 | 1 year ago | |
deeplabv3p | 1 year ago | |
dmnet | 1 year ago | |
dnlnet | 1 year ago | |
emanet | 1 year ago | |
encnet | 1 year ago | |
enet | 1 year ago | |
espnet | 1 year ago | |
espnetv1 | 1 year ago | |
fastfcn | 1 year ago | |
fastscnn | 2 years ago | |
fcn | 1 year ago | |
gcnet | 1 year ago | |
ginet | 1 year ago | |
glore | 1 year ago | |
gscnn | 1 year ago | |
hardnet | 1 year ago | |
hrformer | 1 year ago | |
hrnet_w48_contrast | 1 year ago | |
isanet | 1 year ago | |
knet | 1 year ago | |
lpsnet | 1 year ago | |
lraspp | 1 year ago | |
maskformer | 1 year ago | |
mobileseg | 1 year ago | |
mscale_ocrnet | 1 year ago | |
ocrnet | 1 year ago | |
pfpn | 1 year ago | |
pointrend | 1 year ago | |
portraitnet | 1 year ago | |
pp_humanseg_lite | 1 year ago | |
pp_liteseg | 10 months ago | |
pp_mobileseg | 1 year ago | |
pspnet | 1 year ago | |
pssl | 1 year ago | |
quick_start | 1 year ago | |
rtformer | 1 year ago | |
segformer | 1 year ago | |
segmenter | 2 years ago | |
segnet | 1 year ago | |
segnext | 1 year ago | |
setr | 3 years ago | |
sfnet | 1 year ago | |
smrt | 1 year ago | |
stdcseg | 1 year ago | |
topformer | 1 year ago | |
u2net | 3 years ago | |
uhrnet | 1 year ago | |
unet | 1 year ago | |
unet_3plus | 3 years ago | |
unet_plusplus | 3 years ago | |
upernet | 1 year ago | |
vit_adapter | 1 year ago | |
README.md | 2 years ago | |
README_cn.md | 2 years ago |
English | 简体中文
The config files of different models are saved in PaddleSeg/configs
.
PaddleSeg use the config files to train, validate and export models.
Training datasset
- parameter
- type: Dataset type, please refer to the training configuration file for more details of supported values
- others: Please refer to the corresponding model training configuration file
Evaluation dataset
- parameter
- type: Dataset type, please refer to the training configuration file for more details of supported values
- others: Please refer to the corresponding model training configuration file
On a single card, the amount of data during each iteration of training
Training steps
Training optimizer
- parameter
- type : supports all official optimizers of PaddlePaddle
- weight_decay : L2 regularization value
- others : Please refer to Optimizer
Learning rate
- parameter
- type : learning rate type, supports 10 strategies, namely 'PolynomialDecay', 'PiecewiseDecay', 'StepDecay', 'CosineAnnealingDecay', 'ExponentialDecay', 'InverseTimeDecay', 'LinearWarmup', 'MultiStepDecay', 'NaturalExpDecay', 'NoamDecay'.
- others : Please refer to Paddle official LRScheduler document
Learning rate
- parameter
- value: initial learning rate value
- decay: decay configuration
- type: attenuation type, currently only supports poly
- power: attenuation rate
- end_lr: final learning rate
Loss function
- parameter
- types: list of loss functions
- type: Loss function type, please refer to the loss function library for more details
- ignore_index : The category that needs to be ignored during the training process. The default value is the same train_datasetas ignore_index. It is recommended not to set this item . If you set this, "ignore_index" in loss and train_datasetthe must be the same.
- coef : a list of coefficients corresponding to corresponding loss functions
Model to be trained
- parameter
- type : model type, please refer to the model library for the more details
- others: Please refer to the corresponding model training configuration file
Model export configuration
- parameter
- transforms: Preprocessing operations during prediction. The transforms are the same as train_dataset, val_datasetetc. If you do not fill in this item, the data will be normalized by default.
For more details, please refer to detailed configuration file
No Description
Python Markdown Text Shell Java other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》