Image classification Networks Updating #979

BigFishMaster · 2018-06-11T11:57:27Z

resolve #978

… image_classification

sfraczek

This is not a complete review, I concentrated on some things relevant for me.

sfraczek · 2018-06-12T07:47:39Z

fluid/image_classification/README.md

-
-# unzip all classes data using unzip.sh
-sh unzip.sh
+wget paddl_imagenet2012_dataset_url/ImageNet2012_dataset.tar


Please provide the url

I replace the URL by the official URL and I use download_imagenet2012.sh to get data.

sfraczek · 2018-06-12T07:47:55Z

fluid/image_classification/README.md

-2. Download training and validation label files from [ImageNet2012 url](https://pan.baidu.com/s/1Y6BCo0nmxsm_FsEqmx2hKQ)(password:```wx99```). Untar it into workspace ```ILSVRC2012/```. The files include
+**step-2:** Download training and validation label files
+```
+wget paddl_imagenet2012_label_url/ImageNet2012_label.tar


I update the URL in download_imagenet2012.sh

sfraczek · 2018-06-12T07:53:18Z

fluid/image_classification/README.md


+Models are trained by starting with learning rate ```0.1``` and decaying it by ```0.1``` after each ```30``` epoches, if not special introduced. Available top-1/top-5 validation accuracy on ImageNet 2012 is listed in table. Pretrained models can be downloaded by clicking related model names.


URLs are all pointing to image_classification directory. Clicking doesn't download them.

I cancel the urls without models, and add right urls to networks with trained models

sfraczek · 2018-06-12T07:56:38Z

fluid/image_classification/eval.py


 parser = argparse.ArgumentParser(description=__doc__)
 add_arg = functools.partial(add_arguments, argparser=parser)
-# yapf: disable
-add_arg('batch_size',       int,   32,        "Minibatch size.")
-add_arg('use_gpu',          bool,  True,      "Whether to use GPU or not.")


At intel we are optimizing the CPU path performance of PaddlePaddle. Please do not remove the CPU support.

I add yapf 'disable' and 'enable'

sfraczek · 2018-06-12T08:00:10Z

fluid/image_classification/eval.py


 parser = argparse.ArgumentParser(description=__doc__)
 add_arg = functools.partial(add_arguments, argparser=parser)
-# yapf: disable


Removing the block markers # yapf: disable and # yapf: enable has made this parameters defining section hard to read. I think it was better previously.

Yeah, I agree.

I add yapf 'disable' and 'enable'

sfraczek · 2018-06-12T08:11:32Z

fluid/image_classification/eval.py

+        test_info[1].append(acc1 * len(data))
+        test_info[2].append(acc5 * len(data))
+        cnt += len(data)
+        print("total_number:", cnt)


Someone running this script wouldn't know what is this total_number representing. How about a more self-explanatory message?

I delete this line

sfraczek · 2018-06-12T09:11:28Z

fluid/image_classification/infer.py


 parser = argparse.ArgumentParser(description=__doc__)
 add_arg = functools.partial(add_arguments, argparser=parser)
-# yapf: disable


same as previously.

I fix this.

ok. can you add yapf: disable/enable here too?

sfraczek · 2018-06-12T09:13:17Z

fluid/image_classification/infer.py

+    # model definition
+    model = models.__dict__[model_name]()
+
+    if model_name in ["GoogleNet"]:


same as previously

I update this as you said

sfraczek · 2018-06-12T09:15:34Z

fluid/image_classification/infer.py


-    place = fluid.CUDAPlace(0) if args.use_gpu else fluid.CPUPlace()


same as previously

I fix this.

sfraczek · 2018-06-12T09:23:30Z

fluid/image_classification/train.py

+    # model definition
+    model = models.__dict__[model_name]()
+
+    if model_name in ["GoogleNet"]:


same as previously

I update this as you said

qingqing01

Greate work.

qingqing01 · 2018-06-13T03:07:15Z

fluid/image_classification/README.md


-To start a training task, one can use command line as:
+After data preparation, one can start  the training by:


Remove the extra blank between start and the.

I delete the space

qingqing01 · 2018-06-13T03:09:45Z

fluid/image_classification/eval.py


 parser = argparse.ArgumentParser(description=__doc__)
 add_arg = functools.partial(add_arguments, argparser=parser)
-# yapf: disable


Yeah, I agree.

qingqing01 · 2018-06-13T03:19:40Z

fluid/image_classification/models/dpn.py

+
+        return netArg
+
+    def DualPathFactory(self,


The function name should be dual_path_factory in Google Python style. https://google.github.io/styleguide/pyguide.html

Same as other files and other variables.

I change DualPathFactory to dual_path_factory and revise others which have same problem

qingqing01 · 2018-06-13T03:21:21Z

fluid/image_classification/models/dpn.py

+        }
+        netArg['init_num_filter'] = init_num_filter
+        netArg['init_filter_size'] = init_filter_size
+        netArg['init_padding'] = init_padding


https://google.github.io/styleguide/pyguide.html

netArg -> net_arg. Please follow the Google Python style.

…er/models into image_classification

BigFishMaster · 2018-06-13T11:16:35Z

@sfraczek Hi, I have fix the problems you have metioned. Please review this PR.

qingqing01 · 2018-06-14T09:34:51Z

fluid/image_classification/README.md

+
+## Finetuning
+
+Finetuning is to finetune model weights in a specific task by loading pretrained weights. After initializing ```path_to_pretrain_model``` , one can finetune a model as:
 ```


Need to give the training results here.

qingqing01 · 2018-06-14T09:36:35Z

fluid/image_classification/README_cn.md

+
+## 安装
+
+在当前目录下运行样例代码需要PadddlePaddle的v0.10.0或以上的版本。如果你的运行环境中的PaddlePaddle低于此版本，请根据[安装文档](http://www.paddlepaddle.org/docs/develop/documentation/zh/build_and_install/pip_install_cn.html)中的说明来更新PaddlePaddle。


PadddlePaddle的v0.10.0 -> PadddlePaddle Fluid v0.13.0

qingqing01 · 2018-06-14T09:37:01Z

fluid/image_classification/README.md


-This model built with paddle fluid is still under active development and is not
-the final version. We welcome feedbacks.
+Running sample code in this directory requires PaddelPaddle v0.10.0 and later. If the PaddlePaddle on your device is lower than this version, please follow the instructions in [installation document](http://www.paddlepaddle.org/docs/develop/documentation/zh/build_and_install/pip_install_cn.html) and make an update.


PaddelPaddle v0.10.0 -> PaddelPaddle Fluid v0.13.0

qingqing01 · 2018-06-14T09:38:05Z

fluid/image_classification/README_cn.md

+
+**步骤一：** 从ImageNet官网下载ImageNet-2012的图像数据。训练以及验证数据集会分别被下载到"train" 和 "val" 目录中。
+
+**步骤二：** 下载训练与验证集合对应的标签文件。下面两个文件分别包含了训练集合与验证集合中图像的标签：


告诉用户需要注册ImageNet官网，需要获取key？提醒下用户下载ImageNet时间较长，如果用户已经有了数据，按照下面方式组织。

同步改下英文。

qingqing01 · 2018-06-14T09:38:48Z

fluid/image_classification/data/ILSVRC2012/download_imagenet2012.sh

+
+echo "Download imagenet validation data..."
+mkdir -p ${valid_folder}
+#wget -nd -c ${root_url}/${valid_tar}


qingqing01 · 2018-06-14T09:39:02Z

fluid/image_classification/data/ILSVRC2012/download_imagenet2012.sh

+echo "Download imagenet label file: val_list.txt & train_list.txt"
+label_file=ImageNet_label.tgz
+label_url=http://imagenet-data.bj.bcebos.com/${label_file}
+#wget -nd -c ${label_url}


qingqing01 · 2018-06-14T09:40:55Z

fluid/image_classification/README_cn.md

+       --image_shape=3,224,224 \
+       --with_mem_opt=True \
+       --pretrained_model=${path_to_pretrain_model}
+```


给个示例，预测结果长啥样。

qingqing01 · 2018-06-14T09:41:38Z

fluid/image_classification/README_cn.md

+       --image_shape=3,224,224 \
+       --with_mem_opt=True \
+       --pretrained_model=${path_to_pretrain_model}
+```


给个示例，下载一个已经训练好的模型，这里eval，输出结果长啥样？

…er/models into image_classification

yancanxiang and others added 21 commits May 31, 2018 17:52

small fix.

2d09862

small fix.

a6a3896

add models folder

6385127

update readme

f3bcc11

Update README.md

a394d72

Update README.md

4693b4d

Update README.md

eb89a6e

Update README.md

82bea3e

Update README.md

2dc67ba

Update README.md

7f61a69

Update README.md

0f534bf

Update README.md

9dfcd60

Update README.md

1b301e1

Update README.md

4814877

Update README.md

b1c78e3

Update README.md

604b72c

add dpn

f8131a9

Merge branch 'develop' of https://github.com/PaddlePaddle/models into…

f176173

… image_classification

update train.py

7a6ad78

update train.py

177919f

update train and eval

49af8fd

$sfraczek$

sfraczek suggested changes Jun 12, 2018

View reviewed changes

BigFishMaster and others added 3 commits June 12, 2018 17:56

Update README.md

dfcf841

Update README.md

90712cb

format

4f82ea4

qingqing01 reviewed Jun 13, 2018

View reviewed changes

BigFishMaster and others added 4 commits June 13, 2018 11:28

Merge branch 'image_classification' of https://github.com/BigFishMast…

d21d3c3

…er/models into image_classification

Update README.md

b5feaeb

Update README.md

fc71934

update

339af6f

BigFishMaster and others added 6 commits June 13, 2018 17:23

Merge branch 'image_classification' of https://github.com/BigFishMast…

01a7b89

…er/models into image_classification

Update README.md

6908903

update shell

1835b1f

Merge branch 'image_classification' of https://github.com/BigFishMast…

299fd41

…er/models into image_classification

Update README.md

609b87f

add yapf disable to args

96661c5

BigFishMaster added 2 commits June 13, 2018 19:27

Create README_cn.md

020f94b

Update README_cn.md

6bd9871

$sfraczek$

sfraczek approved these changes Jun 13, 2018

View reviewed changes

update googlenet

a744622

qingqing01 reviewed Jun 14, 2018

View reviewed changes

BigFishMaster and others added 16 commits June 14, 2018 19:17

add images folder

bb11061

Update README.md

5d5a1df

Update README.md

df07ee4

add curve.jpg

0e44ef4

Merge branch 'image_classification' of https://github.com/BigFishMast…

de50be2

…er/models into image_classification

Update README.md

4b611d0

update download_imagenet2012.sh

b9c62e2

Update README.md

8db82cb

Merge branch 'image_classification' of https://github.com/BigFishMast…

1f6005f

…er/models into image_classification

Update README.md

a6e9f0f

Update README.md

8113cbb

Update README.md

94aaea6

Update README.md

e8d2db6

Update README_cn.md

3803d34

Update README.md

2c59d9e

Update README_cn.md

20080a7

qingqing01 approved these changes Jun 15, 2018

View reviewed changes

qingqing01 merged commit 843a00f into PaddlePaddle:develop Jun 15, 2018


		Models are trained by starting with learning rate ```0.1``` and decaying it by ```0.1``` after each ```30``` epoches, if not special introduced. Available top-1/top-5 validation accuracy on ImageNet 2012 is listed in table. Pretrained models can be downloaded by clicking related model names.


		place = fluid.CUDAPlace(0) if args.use_gpu else fluid.CPUPlace()


		To start a training task, one can use command line as:
		After data preparation, one can start the training by:


		## 安装

		在当前目录下运行样例代码需要PadddlePaddle的v0.10.0或以上的版本。如果你的运行环境中的PaddlePaddle低于此版本，请根据[安装文档](http://www.paddlepaddle.org/docs/develop/documentation/zh/build_and_install/pip_install_cn.html)中的说明来更新PaddlePaddle。


		步骤一：从ImageNet官网下载ImageNet-2012的图像数据。训练以及验证数据集会分别被下载到"train" 和 "val" 目录中。

		步骤二：下载训练与验证集合对应的标签文件。下面两个文件分别包含了训练集合与验证集合中图像的标签：

Image classification Networks Updating #979

Image classification Networks Updating #979

Conversation

BigFishMaster commented Jun 11, 2018 • edited Loading

sfraczek left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BigFishMaster Jun 13, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

qingqing01 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BigFishMaster commented Jun 13, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BigFishMaster commented Jun 11, 2018 •

edited

Loading

$@sfraczek$ sfraczek left a comment

BigFishMaster Jun 13, 2018 •

edited

Loading