Object Detection

This is an implementation of SSD for object detection in Tensorflow. It contains complete code for preprocessing, postprocessing, training and test. Besides, this repository is easy-to-use and can be developed on Linux and Windows.

SSD : Liu, Wei, et al. "Ssd: Single shot multibox detector." European conference on computer vision. Springer, Cham, 2016.

Getting Started

1 Prerequisites

• Python 3.x
• Tensorflow 1.x
• Opencv-python
• Pandas

2 Define your class names

Download and unzip this repository.
cd ../SSD/label
Open the label.txt and revise its class names as yours.

3 Prepare images

Copy your images and annotation files to directories ../SSD/data/annotation/images and ../SSD/data/annotation/images/xml respectively, where the annotations should be obtained by a graphical image annotation tool and saved as XML files in PASCAL VOC format.
cd ../SSD/code
run
python spilt.py
Then train and val images will be generated in ../SSD/data/annotation/train and /SSD/data/annotation/test directories, respectively.

4 Generate anchors (default boxes)

cd ../SSD/code
run
python anchor_generators.py
Anchors generated will be saved in the directory ../SSD/anchor/anchor.txt.

5 Train model using Tensorflow

The model parameters, training parameters and eval parameters are all defined by parameters.py.
cd ../SSD/code
run
python train.py
The model will be saved in directory ../SSD/model/checkpoint, and some detection results are saved in ../SSD/pic.

6 Visualize model using Tensorboard

cd ../SSD
run
tensorboard --logdir=model/
Open the URL in browser to visualize graph of the model, as follows:

Local Graph

Examples

Belows are some successful detection examples in my dataset: