AIML Service

Integrate ONNX Runtime on the West Connectivity for Deep Learning Models. Visit ONNX support to learn more on ONNX supported AI model conversions.

Operations

Aiml.getAsyncImageInference() - Returns an inference result based of an image as an aiml event.
Aiml.getImageInference() - Returns an inference result based of an image.
Aiml.inference() - Return an inference result based on input dataset.
Aiml.modelSignature() - Return a model signature based on the provided model name.

Events

event - An event triggered by sending specific requests to aiml. Configurable on Config UI.

Operations

Please note that bold arguments are required, while italic arguments are optional.

getAsyncImageInference

The input includes an image and onnx ai model name stored on a solution's content-service. The service also provide basic pre-processing steps to the image before prediction.

Arguments

parameters (object) - Object containing service call parameters.

Name	Type	Description
model_name	string	The name of a model.
image_name	string	The name of a model.
pre_process	object	Pre-processing steps to the image before the model generates inference result
pre_process.order	string	Comma separated string, to allow aiml service to execute pre-processing steps in the given order. Values can include allow_grayscale_processing, crop_range, allow_center_crop, allow_normalization, allow_resize
pre_process.crop_range	string	Four comma separated numbers(left, top, right, bottom), to allow aiml to crop the image. If all numbers are between 0 and 1, numbers become the percentage of the image.
pre_process.allow_resize	boolean	Allow aiml service to resize image size to match model signature.
pre_process.allow_center_crop	boolean	Allow aiml service to apply default square cropping pre-processing steps to the image.
pre_process.allow_normalization	boolean	Allow aiml service to apply image normalization to the image.
pre_process.allow_grayscale_processing	boolean	Allow aiml service to first convert the image into grayscale and then convert the image into BGR format.
post_process	object	Post-processing steps to the inference result
post_process.hide_outputs	string	Comma separated values to filter model inference output. Useful if the output shows Exceeded allotted memory! error or if the output take too long to process. Do not include this in order

Responses

Code	Body Type	Description
200	string	The acknowledgement receipt of the request. Fixed value "OK".
429	object	Too Many Requests. Please re-try in 1 minute or check rate-limit settings.
500	object	Internal Server Error

Object Parameter of 429 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 500 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Example

local post_process = {}
  -- post_process = {hide_outputs="Output1,Output6,Output7,Output8,Output4"}
local pre_process = {} pre_process = {
  crop_range="0,0,1,1",
  allow_normalization = true
  -- order ="crop_range,allow_normalization"
} return Aiml.getAsyncImageInference({
  model_name ="mobilenet_v2_roi_topbottom_v3_1.onnx",
  image_name ="222203025_C51_R_Sidewall_11_ToolSet_Image1__000024", 
  pre_process = pre_process, 
  post_process = post_process
})

getImageInference

The input includes an image and onnx ai model name stored on a solution's content-service. The service also provide basic pre-processing steps to the image before prediction. Please increase cache size if you are experiencing long inference wait time and use many models.

Arguments

parameters (object) - Object containing service call parameters.

Name	Type	Description
model_name	string	The name of a model.
image_name	string	The name of a model.
pre_process	object	Pre-processing steps to the image before the model generates inference result
pre_process.order	string	Comma separated string, to allow aiml service to execute pre-processing steps in the given order. Values can include allow_grayscale_processing, crop_range, allow_center_crop, allow_normalization, allow_resize
pre_process.crop_range	string	Four comma separated numbers(left, top, right, bottom), to allow aiml to crop the image. If all numbers are between 0 and 1, numbers become the percentage of the image.
pre_process.allow_resize	boolean	Allow aiml service to resize image size to match model signature.
pre_process.allow_center_crop	boolean	Allow aiml service to apply default square cropping pre-processing steps to the image.
pre_process.allow_normalization	boolean	Allow aiml service to apply image normalization to the image.
pre_process.allow_grayscale_processing	boolean	Allow aiml service to first convert the image into grayscale and then convert the image into BGR format.
post_process	object	Post-processing steps to the inference result
post_process.hide_outputs	string	Comma separated values to filter model inference output. Useful if the output shows Exceeded allotted memory! error or if the output take too long to process. Do not include this in order

Responses

Code	Body Type	Description
200	object	The inference result
400	object	Bad Request
422	object	Unprocessable Content
429	object	Too Many Requests. Please re-try in 1 minute or check rate-limit settings.
500	object	Internal Server Error

Object Parameter of 200 response:

Name	Type	Description
OutputN	object	Inference result output. Starting from Output1 to OutputN depending on the model

Object Parameter of 400 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 422 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 429 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 500 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Example

local post_process = {}
  -- post_process = { hide_outputs = "Output2"}
local pre_process = {} pre_process = {
  crop_range="0,0,1,1",
  allow_normalization = true
  -- order ="crop_range,allow_normalization"
} return Aiml.getImageInference({model_name="mobilenet_v2_roi_topbottom_v3_1.onnx",
  image_name="222203025_C51_R_Sidewall_11_ToolSet_Image1__000024", pre_process = pre_process, 
  post_process=post_process
})

inference

Return an inference result based on input dataset.

Arguments

parameters (object) - Object containing service call parameters.

Name	Type	Description
dataset	[ [ number ] ]	The numerical dataset, should be a 2D array.
model_name	string	The name of a model.

Responses

Code	Body Type	Description
200	object	The inference result
400	object	Bad Request
500	object	Internal Server Error

Object Parameter of 200 response:

Name	Type	Description
OutputN	object	Inference result output. Starting from Output1 to OutputN depending on the model

Object Parameter of 400 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 500 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Example

Aiml.inference({
  dataset = {
    {7, 3.2 , 4.7, 1.4}
  },
  model_name = "logreg_iris.onnx"
})



Aiml.inference({
  dataset = {
    {7, 3.2 , 4.7, 1.4},
    {6.3, 2.3, 4.4, 1.3},
    {5.1, 3.7 , 1.5, 0.4}
  },
  model_name = "logreg_iris.onnx"
})

modelSignature

Return a model signature based on the provided model name.

Arguments

parameters (object) - Object containing service call parameters.

Name	Type	Description
model_name	string	The name of a model.

Responses

Code	Body Type	Description
200	object	The model signature
400	object	Bad Request
500	object	Internal Server Error

Object Parameter of 200 response:

Name	Type	Description
inputs	object	The input fields and format
outputs	object	The output fields and format
filename	string	The model file name
createtime	string	The creation date time of model signature, in ISO format
sess_options	object	The Onnx Runtime session options
sess_options.execution_mode	string	The Onnx Runtime parameter to control whether to execute operators in the graph sequentially or in parallel
sess_options.intra_op_num_threads	string	The Onnx Runtime parameter to control the total number of INTRA threads to use to run the model
sess_options.graph_optimization_level	string	The Onnx Runtime parameter to determine optimization level

Object Parameter of 400 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Object Parameter of 500 response:

Name	Type	Description
type	string	Error type
error	string	Error message
status	integer	Response code

Example

Aiml.modelSignature({model_name = "logreg_iris.onnx"})

Events

event

An event triggered by sending specific requests such as getAsyncImageInference. Configurable on Config UI.

Arguments

event (object) - A prediciton result for asynchronous inference.

Name	Type	Description
output	object	Prediction result. Starting from Output1 to OutputN depending on the model
source	object	Source model, image, and camera_ip address.
end_time	integer	Request end timestamp in milliseconds
start_time	integer	Request start timestamp in milliseconds, 0 when undefined.

Example

--updates solution env on timer timeout
Config.updateEnv({
  aiml_start_time = "aiml event started at " .. event.start_time
})
Config.updateEnv({
  aiml_end_time = "aiml event ended at " .. event.end_time
})
Config.updateEnv({
  aiml_source = "aiml source contains " .. event.source
})