vsdkx-cli is a cli tool to facilitate the usage of vsdkx. you can add/remove any vsdkx addons or vsdkx models to your project.

How to install

To install this client you can easily use pip.

pip install git+https://github.com/natix-io/vsdkx-core
pip install git+https://github.com/natix-io/vsdkx-cli

After this you can use vsdkx-cli which is installed in the PATH

Usage

To install vsdkx models you can use easily go to your project and run

vsdkx-cli model add {model_name} {weight_name}
#or to remove any model you can use remove instead of add

currently these models and weights are available

Model	Weights
yolo-torch	default.pt, best.pt
yolo-tflite	default.pt, best.pt
yolo-facemask	default.pt, best.pt
bayesian	default.pt, best.pt
resnet	default.pt, best.pt
mobilenet	default.pt, best.pt
To add any addon to your project you can use:

vsdkx-cli addon add {addon_name}
#or to remove any addon you can use remove instead of add

currently these addons are available:

• zoning
• distant
• tracking

To set new configuration for the repo that vsdk-cli is searching for the models and addons you can use:

vsdkx-cli repo

Notice: You should run all the vsdkx-cli command inside your project directory

To clean all the things you have done with this cli just do this:

vsdkx-cli clean

---

After everything is installed inside your project now you should have these directories inside your project

|-- project
|   |-- vsdkx
|       |-- model
|           |-- profile.yaml
|       |-- weight
|           |-- weight files according to your models

the profile.yaml contains information about the model and which model uses which weight file. you can either manually change that file or you can use:

vsdkx-cli set {model_name} {weight_name}

Now for final step you only need to config the app with this:

vsdkx-cli app init

This will create settings.yaml file and vsdkx-run.py inside your project and in the vsdkx folder according the models and addons you added before. you can also change the addons and models with previous commands but if you want to update settings.yaml with new models and addons you can reconfigure the app with:

vsdkx-cli app config

now to use the vsdkx-run.py you can run python vsdkx-run.py -h to see the usage

You can also configure the drawing properties for debug mode by using:

vsdkx-cli app draw

You can see the list of models and addons in you project by using:

vsdkx-cli app list

You can either use SimpleRunner which can bring up the application and you can pass required settings through gRPC or command line arguments.

from vsdkx.core.runner import SimpleRunner

SimpleRunner().run()

or you can import EventDetector manually and pass the required settings to its constructor.

from vsdkx.core.util import io
from vsdkx.core.detector import EventDetector

detector = EventDetector(io.import_yaml("Path to settings.yaml"))
result = detector.detect(image)

You can see the Settings section to see the structure of the configurations.

If you prefer to use SimpleRunner you can pass a function to its constructor to get the result of the model that way.

from vsdkx.core.runner import SimpleRunner

SimpleRunner(lambda image, inference: {
    # do anything you want with inference.
}).run()

Hidden Files

This cli will create two hidden files .secret which has the credential information for the registry that we are using to download weight files and other data, and .config file which stores the list of addons and model-drivers for this project.

Profile

profile.yaml is a file inside visionx/model and we can configure the model that we are going to use with its attributes. it should have following structure.

profile-name:
# All the required parameter for that specific model
  classes_len: 
  input_shape:
  model_path: # path to weight file
  ...
    

We can have multiple profiles and we can pass profile-name to the driver to use that specific values.

Settings

settings.yaml is a file inside visionx directory and has following structure.

addons:
    addon-name:
      class:  # the class of the addon, each addon has a specific class, it will be added automatically if you use cli to add addons to your project
      ... # other properties required by that addon like distance_threshold, max_disappeared, etc that we need to use for tracking addon
model:
    class: # this is the class of the model driver we want to use for this project, it will be added automatically if you use cli to add model-driver to your project
    debug: # If this value is true it will show debug window with some simple drawings, like zonings, bboxes and etc.
    profile: # the name of the profile we want to use for our model driver
    settings:
      ... # all other dynamic settings the we can pass to our model driver
drawing: # this is optional if we want to have debug window with minimal drawing feature, this dictionary will also be passed to addons and model-drivers so if developers need to debug something there they can set drawing configs here.
  box_font_scale: 0.8
  box_thickness: 3
  group_color: !!python/tuple
  - 135
  - 206
  - 235
  rectangle_color: !!python/tuple
  - 60
  - 179
  - 113
  text_color: !!python/tuple
  - 255
  - 255
  - 255
  text_fontscale: 1
  text_thickness: 1
  zone_thickness: 3
  zones:
  - - - 538
      - 226
    - - 837
      - 213
    - - 877
      - 656
    - - 589
      - 674
    - - 538
      - 226
  zones_color: !!python/tuple
  - 153
  - 50
  - 204

# we can pass this yaml file through gRPC to our project or via command line