From 022578f7812536a1922ed4179419ea52b0aefb88 Mon Sep 17 00:00:00 2001 From: root Date: Tue, 15 Jan 2019 13:41:46 +0000 Subject: [PATCH] fixed links and docs --- fluid/PaddleCV/deeplabv3+/README.md | 2 +- fluid/PaddleCV/face_detection/README_cn.md | 2 +- fluid/PaddleNLP/text_classification/README.md | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/fluid/PaddleCV/deeplabv3+/README.md b/fluid/PaddleCV/deeplabv3+/README.md index 9ff68ab8..97e1600d 100644 --- a/fluid/PaddleCV/deeplabv3+/README.md +++ b/fluid/PaddleCV/deeplabv3+/README.md @@ -76,7 +76,7 @@ python ./train.py \ --train_crop_size=769 \ --total_step=90000 \ --init_weights_path=deeplabv3plus_xception65_initialize.params \ - --save_weights_path=output \ + --save_weights_path=output/ \ --dataset_path=$DATASET_PATH ``` diff --git a/fluid/PaddleCV/face_detection/README_cn.md b/fluid/PaddleCV/face_detection/README_cn.md index 80485009..f63fbed0 100644 --- a/fluid/PaddleCV/face_detection/README_cn.md +++ b/fluid/PaddleCV/face_detection/README_cn.md @@ -99,7 +99,7 @@ python -u train.py --batch_size=16 --pretrained_model=vgg_ilsvrc_16_fc_reduced 模型训练所采用的数据增强: -**数据增强**:数据的读取行为定义在 `reader.py` 中,所有的图片都会被缩放到640x640。在训练时还会对图片进行数据增强,包括随机扰动、翻转、裁剪等,和[物体检测SSD算法](https://github.com/PaddlePaddle/models/blob/develop/fluid/object_detection/README_cn.md#%E8%AE%AD%E7%BB%83-pascal-voc-%E6%95%B0%E6%8D%AE%E9%9B%86)中数据增强类似,除此之外,增加了上面提到的Data-anchor-sampling: +**数据增强**:数据的读取行为定义在 `reader.py` 中,所有的图片都会被缩放到640x640。在训练时还会对图片进行数据增强,包括随机扰动、翻转、裁剪等,和[物体检测SSD算法](https://github.com/PaddlePaddle/models/blob/develop/fluid/PaddleCV/object_detection/README.md)中数据增强类似,除此之外,增加了上面提到的Data-anchor-sampling: **尺度变换(Data-anchor-sampling)**:随机将图片尺度变换到一定范围的尺度,大大增强人脸的尺度变化。具体操作为根据随机选择的人脸高(height)和宽(width),得到$v=\\sqrt{width * height}$,判断$v$的值位于缩放区间$[16,32,64,128,256,512]$中的的哪一个。假设$v=45$,则选定$32