# ASOT(无约束交通场景自动分析系统)
*--Automatic analysis system of unconstrained traffic scene*

## Dependencies(环境需求)
* Python 3
* PyTorch >= 1.1.0
* NVIDIA GPU + CUDA
## Python packages（安装库）
* pip install（待续）

## Preparation(准备)
### [Get the code](https://github.com/yukichou/transport_ana)

git clone https://github.com/yukichou/transport_ana.git
or download zip transport_ana-master.zip & unzip
cd ./transport_ana-master

### Download the .pth file
* First you need to download the checkpoints.zip from the Baidu Drive(extraction code:1234) and unzip
* the transport_ana file store pth in a file structure like this:

   transport_ana
   └───car_class
    │    └─checkpoints   
    │       └──epoch_39.pth
    └───reid
    │    └─backup
    │       └─resnet50_80.pth
    └───config
        │ yolov3.weights
        └-yolov3.weights.1

* you need to follow the file structure and move the .pth file to the right place
## Usage（使用说明）
### CAM mod
1. To use this system by compiling the python file:

python final_ui.py

2. Then a surface will raise ,it looks like below:

3. Before you push the CAM mod button , please use the File-open button which on the left-top corner to load the video you want to analyse.

4. After loading the video , pushing the CAM button then it just works.

### REID mod
1. The button besides the CAM mod is the REID mod (Vehicle RE-identification)
2. Load the car image you want to search and push the REID button , a sub-win show up :
3. Cars from the gallery set are extracted from the previous video, after matching calculation, you will get the the most similar car that compare to the query car.

###### tags: `yuki`