Deep Learning (DL) · Advanced level

Advanced level · Module 4: Deploying Deep Learning Models · Lesson 3 of 4

Offline Deployment on Phones

Run models directly on phones so they work without internet, and measure real performance.

In this lesson you will

  • Run a model in a mobile or edge runtime
  • Measure on-device latency, memory and battery effects

Key concepts

on-device inferencemobile runtimeslatency and memoryoffline-first design

The lesson

On-device inference keeps working without connectivity, keeps photos on the phone (better privacy) and avoids data costs — important for field work in remote areas.

Mobile runtimes such as ONNX Runtime Mobile, PyTorch's mobile options or TensorFlow Lite run exported models on Android. Test on a realistic mid-range phone, not a flagship.

Measure latency per prediction, memory use and battery impact. Plan how the app receives model updates when a connection is available.

Step by step

  1. Load the compressed model in a mobile or edge runtime.
  2. Run predictions on sample photos offline.
  3. Measure latency and memory.
  4. Plan model updates over intermittent connectivity.

Practice activity

Demonstrate the project model running offline on a phone or edge device.

Expected result: A working offline demonstration with measured latency and memory.

Check your understanding

  1. Name two benefits of on-device inference in remote areas.

    Show answer

    It works offline and keeps data on the device (also avoiding data costs).

  2. Why test on a mid-range phone?

    Show answer

    It reflects the devices field users actually have.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.