Engine.PerformanceTest

Run a performance test on the Engine to determine prediction times and how well a model is performing.

PerformanceTestRequest

feature_type
string

The type of feature to performance test against

class
string

The class to performance test against

image
ImageItem

A sample image used for performance test. There are also no restrictions around image size, these are defined at the app layer. Supported image formats include PNG, JPEG and GIF.

feed_id
string

User specified ID to reference the source of the image

custom_id
string

User specified ID used to identify an image. This optional custom ID can be used to map your image request to a result from Engine. Engine does not perform any operations with custom ID, it merely just passes the value directly back in the response. It is up to the client to ensure uniqueness.

group_id
string

User specified GroupID used to identify an image. This is only used by the Face Recognition feature.

data
string

Either encoded or pixel image data

exclusion_zones
array[ExclusionZone]

Zones to ignore on this image (optional)

points
array[Point]

Ordered list of points delimitating this polygon

x
int32

Position on the X-axis

y
int32

Position on the Y-axis

PerformanceTestResponse

batch_sizes
array[BatchSizesEntry]

A map of batch sizes to their performance. Engine will run a performance test on varying batch sizes and return the performance characteristics for each one. This will help give an idea on how a model will perform on some given hardware.

key
int32
value
BatchSizePerformanceResponse
average_image_prediction_time
double

Average amount of time taken to make a prediction

total_prediction_time
double

Total amount of time taken to make a prediction