xxq250
/
mindspore
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

add readme

This commit is contained in:
unknown 2020-05-29 12:00:10 +08:00
parent eef3c58b5e
commit e66b23b357
1 changed files with 31 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
model_zoo/deeplabv3/README.md
View File

@ -1,39 +1,42 @@
Deeplab-V3 Example # Deeplab-V3 Example
Description ## Description
This is an example of training DeepLabv3 with PASCAL VOC 2012 dataset in MindSpore. This is an example of training DeepLabv3 with PASCAL VOC 2012 dataset in MindSpore.
Paper Rethinking Atrous Convolution for Semantic Image Segmentation Paper Rethinking Atrous Convolution for Semantic Image Segmentation
Liang-Chieh Chen, George Papandreou, Florian Schroff, Hartwig Adam Liang-Chieh Chen, George Papandreou, Florian Schroff, Hartwig Adam
Requirements ## Requirements
Install MindSpore. - Install [MindSpore](https://www.mindspore.cn/install/en).
Download the VOC 2012 dataset for training. - Download the VOC 2012 dataset for training.
For more information, please check the resources below：
MindSpore tutorials
MindSpore API
Notes: If you are running a fine-tuning or evaluation task, prepare the corresponding checkpoint file. > Notes:
If you are running a fine-tuning or evaluation task, prepare the corresponding checkpoint file.
Running the Example ## Running the Example
### Training
Training - Set options in config.py.
Set options in config.py. - Run `run_standalone_train.sh` for non-distributed training.
Run run_standalone_train.sh for non-distributed training. ``` bash
sh scripts/run_standalone_train.sh DEVICE_ID EPOCH_SIZE DATA_DIR sh scripts/run_standalone_train.sh DEVICE_ID EPOCH_SIZE DATA_DIR
Run run_distribute_train.sh for distributed training. ```
sh scripts/run_distribute_train.sh DEVICE_NUM EPOCH_SIZE DATA_DIR MINDSPORE_HCCL_CONFIG_PATH - Run `run_distribute_train.sh` for distributed training.
``` bash
Evaluation sh scripts/run_distribute_train.sh DEVICE_NUM EPOCH_SIZE DATA_DIR MINDSPORE_HCCL_CONFIG_PATH
```
### Evaluation
Set options in evaluation_config.py. Make sure the 'data_file' and 'finetune_ckpt' are set to your own path. Set options in evaluation_config.py. Make sure the 'data_file' and 'finetune_ckpt' are set to your own path.
Run run_eval.sh for evaluation. - Run run_eval.sh for evaluation.
sh scripts/run_eval.sh DEVICE_ID DATA_DIR ``` bash
sh scripts/run_eval.sh DEVICE_ID DATA_DIR
```
Options and Parameters ## Options and Parameters
It contains of parameters of Deeplab-V3 model and options for training, which is set in file config.py. It contains of parameters of Deeplab-V3 model and options for training, which is set in file config.py.
Options: ### Options:
```
config.py: config.py:
learning_rate Learning rate, default is 0.0014. learning_rate Learning rate, default is 0.0014.
weight_decay Weight decay, default is 5e-5. weight_decay Weight decay, default is 5e-5.
@ -49,10 +52,11 @@ config.py:
decoder_output_stride The ratio of input to output spatial resolution when employing decoder decoder_output_stride The ratio of input to output spatial resolution when employing decoder
to refine segmentation results, default is None. to refine segmentation results, default is None.
image_pyramid Input scales for multi-scale feature extraction, default is None. image_pyramid Input scales for multi-scale feature extraction, default is None.
```
Parameters: ### Parameters:
```
Parameters for dataset and network: Parameters for dataset and network:
distribute Run distribute, default is false. distribute Run distribute, default is false.
epoch_size Epoch size, default is 6. epoch_size Epoch size, default is 6.
@ -61,4 +65,5 @@ Parameters for dataset and network:
checkpoint_url Checkpoint path, default is None. checkpoint_url Checkpoint path, default is None.
enable_save_ckpt Enable save checkpoint, default is true. enable_save_ckpt Enable save checkpoint, default is true.
save_checkpoint_steps Save checkpoint steps, default is 1000. save_checkpoint_steps Save checkpoint steps, default is 1000.
save_checkpoint_num Save checkpoint numbers, default is 1. save_checkpoint_num Save checkpoint numbers, default is 1.
```