Model the System Before Building It

Compare local and remote inspection time

Send a smaller image crop

The remote inspection model computes quickly, but Botie's part must wait for image upload and network overhead as well.

BOTIE

What allows the remote route to become faster in the small-crop case?
The inputs we are comparing

Image payload in bytes

Large image

600000

Small crop

60000

Effective upload bytes per second

Same in both cases

3000000

Fixed network overhead in seconds

Same in both cases

0.04

Remote computation seconds

Same in both cases

0.03

Local computation seconds

Same in both cases

0.12
Result for large image
0.15
Choose an explanation to test
A small request can still lose to connection overhead

This check uses the pieces from “Compare complete local and remote time estimates”.

BOTIE

Botie compares cloud and local inference latency. The payload is 20,000 bytes and upload bandwidth is 2,000,000 bytes per second. Add fixed network overhead and remote compute time, then subtract local compute time. All supplied times are in seconds.
Image payload in bytes
20000
Effective upload bytes per second
2000000
Fixed network overhead in seconds
0.15
Remote computation seconds
0.02
Local computation seconds
0.1

Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.

Build a construction in the workspace above first. This case will test those pieces.

What would need checking before adopting the smaller crop as the production input?

Compare the entire route and test image quality, queueing, outages, and deadlines beyond this nominal timing estimate.

Shared preparation is excluded equally. Server queueing, response payload, outages, energy, and prediction quality need further evidence.

Work through the Python challenges →
Continue the courseWhich Designs Are Actually Acceptable? →
← Will the Model Fit on the Device?