This is a website api for SegNet, enable users and trainers use our model to do image segmentation and create their own models.
It also provides a base website interface to use our apis.
Of course, you can use our apis in your own application.
Make sure you have installed caffe.
You are also required to install nodejs and npm.
Also, you should make sure your server have at least one powerful gpu to do deep learning. Over than Nvidia M40 is needed.
First, download this project and get some dependences
git clone https://github.com/Woolseyyy/SegNet-WebsiteAPI.git
cd SegNet-WebsiteAPI
npm install
Then, you should config some script files path in utils/config.js, you can read segnet tutorial to understand the scripts meaning.
Also you should set blockSize in that file. It means the least block size your training model need. It is determined by the argument * batch size * in your model
Finally, for base testing, you are required to config your base model, named 0, as following.
+Your Project Root Path
+models
+0
+TempTestInterface
+tempTestResult
-test_weights.caffemodel
test_weights.caffemodel is the weights you pre training. It is userd for those users who just want to conduct a image segmentation. You are recomanded to understand segnet tutorial first, and the test_weights.caffemodel can be seen in the test part.
nodejs bin/www
| api | parameters | description |
|---|---|---|
| /api/test | [Model ID], Files | Upload images and conduct segmentation |
| api | parameters | description |
|---|---|---|
| /api/create | Model password, User name, User password | Upload images and conduct segmentation |
| /api/train/prepare/data | Model ID, Model password, Files | Upload data of datasets |
| /api/train/prepare/label | Model ID, Model password, Files | Upload label of datasets |
| /api/train/prepare/relation | Model ID, Model password, Files | Upload relation of datasets |
| /api/train/start | Model ID, Model password | Start Training |
| /api/train/procedure | Model ID, Query range | Watch procedure of training |
| /api/train/stop | Model ID, Model password | Stop Training |
| /api/train/finish | Model password, User name | Save weights |
| /api/train/clear | Model ID, Model password | Clear training model |