# visualization-vgg16

**Repository Path**: xtudbxk/visualization-vgg16

## Basic Information

- **Project Name**: visualization-vgg16
- **Description**: 使用 visualizing and understanding convolutional networks 论文里提到的 deconv network 对 vgg16 进行可视化
- **Primary Language**: Unknown
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 1
- **Forks**: 0
- **Created**: 2017-07-29
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

lization-vgg16
使用 [visualizing and understanding convolutional networks](https://link.springer.com/chapter/10.1007/978-3-319-10590-1_53) 论文里提到的 deconv network 对 [vgg16](http://ethereon.github.io/netscope/#/preset/vgg-16) 进行可视化

note:
1. the base vgg16 construct source code is from [https://github.com/machrisaa/tensorflow-vgg](https://github.com/machrisaa/tensorflow-vgg)
2. before you start the script file, you should do somethings following:
>1. download the vgg16.npy from [http://pan.baidu.com/s/1jHS95Ym](http://pan.baidu.com/s/1jHS95Ym) for china and [https://mega.nz/#!YU1FWJrA!O1ywiCS2IiOlUCtCpI6HTJOMrneN-Qdv3ywQP5poecM](https://mega.nz/#!YU1FWJrA!O1ywiCS2IiOlUCtCpI6HTJOMrneN-Qdv3ywQP5poecM) for other countries.
>2. mkdir a input dir, and put a image named "test.jpg" for the input image.
>3. mkdir a output dir for the output images.

3. to start the script, you just to type `python visiualization_vgg16.py which_gpu_device_to_use`.
4. if you have no gpus, just comment the `os.environ["CUDA_VISIBLE_DEVICES"]=sys.argv[1]` in the script.

some examples:
input image:
![test.jpg](input/test.jpg)
visualization results:
![pool1_6.png](output/pool1_6.png) ![pool1_11.png](output/pool1_11.png)

![pool2_105.png](output/pool2_105.png) ![pool2_127.png](output/pool2_127.png)