Skip to content
简明 SSD 目标检测模型 keras version(交通标志识别 训练部分见 dev 分支)
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
output Modify Readme Dec 8, 2017
pics first commit Dec 8, 2017
video_utils first commit Dec 8, 2017
LICENSE first commit Dec 8, 2017
README.md Update keras version Jan 23, 2019
SSD.ipynb Modify Readme Dec 8, 2017
SSD_crop.py first commit Dec 8, 2017
SSD_crop.py~ first commit Dec 8, 2017
ssd.py first commit Dec 8, 2017
ssd.pyc first commit Dec 8, 2017
ssd_layers.py first commit Dec 8, 2017
ssd_layers.pyc first commit Dec 8, 2017
ssd_utils.py first commit Dec 8, 2017
ssd_utils.pyc first commit Dec 8, 2017

README.md

Single Shot MultiBox Detector Keras version.

SSD是一种Object Detection方法。本文是基于论文SSD: Single Shot MultiBox Detector,实现的keras版本。

该文章在既保证速度,又要保证精度的情况下,提出了SSD物体检测模型,与现在流行的检测模型一样,将检测过程整个成一个single deep neural network。便于训练与优化,同时提高检测速度。 SSD将输出一系列离散化(discretization)的bounding boxes,这些bounding boxes是在不同层次(layers)上的feature maps上生成的,并且有着不同的aspect ratio。

模型效果

  • 模型对载具的检测

  • 模型对动物的检测

  • 模型的视频检测


如何使用

所需依赖

cv2==3.3.0
keras==2.2.0
matplotlib==2.1.0
tensorflow==1.3.0
numpy==1.13.3

如果想跑通视频模块,则需额外pip install scikit-video

具体操作

git clone git@github.com:kuhung/SSD_keras.git
cd SSD_keras
  • Download model weight weights_SSD300.hdf5here
cp weights_SSD300.hdf5 into SSD_keras
  • 对于图片的检测

参考SSD.ipynb

  • 若要剪切图片为下一步处理做准备

参考SSD_crop.py

  • 检测视频
cd video_utils
python videotest_example.py hy.mp4

参考资料

SSD: Single Shot MultiBox Detector

论文阅读:SSD: Single Shot MultiBox Detector

rykov8/ssd_keras

You can’t perform that action at this time.