Learn · Try · Check
What you’ll learn
Choose a processing location around the real workflow.
Listen to this lesson
Local AI narration · Quiz answers are kept out of the audio.
Read the audio transcript
AI 101. Images, devices and deployment choices. Choose a processing location around the real workflow. Recognizing a pattern is not establishing a cause. Computer vision can help classify an image, locate objects or compare captures. The input is still a measurement of a scene under particular lighting, angle and camera conditions. Two photographs can look different because the environment changed or because the capture conditions changed. An interpretation such as material damage or a biological cause needs evidence beyond a pattern match. Local and cloud processing have tradeoffs. On-device inference can work without a network and keep some processing local. It is constrained by memory, battery, hardware support and model size. Cloud processing can offer different resources but introduces connectivity, data-transfer and service considerations. A hybrid workflow can combine them. “On-device” describes where an operation runs; it does not prove that the entire application never uploads information. Check the whole data path. Measure the experience that matters. Ask how long a result takes on the target device, what happens offline and which uncertain cases go to a person. Test with representative inputs instead of only convenient screenshots. The best model for a benchmark may not be the best fit for a field worker holding a phone in poor light. Preserve the original capture and the reviewer’s decision so later users can understand how a report was produced. HerbDev example. OrganicVision is described as an engineering research platform, with physical validation still an evidence boundary. Use that distinction: a reproducible research workflow is useful, but it is not a claim that every real-world setting has been validated. The linked embedded-AI guide explores local, cloud and hybrid constraints. The process: Capture evidence. Then Choose processing location. Then Measure on target hardware. Then Review uncertain results. Try this. A field assistant must work with no signal. Name two questions you would ask before choosing a cloud-only design. Pause the audio to try the exercise. A worked answer and checkpoint questions are available on the lesson page.
Recognizing a pattern is not establishing a cause
Computer vision can help classify an image, locate objects or compare captures. The input is still a measurement of a scene under particular lighting, angle and camera conditions. Two photographs can look different because the environment changed or because the capture conditions changed. An interpretation such as material damage or a biological cause needs evidence beyond a pattern match.
Local and cloud processing have tradeoffs
On-device inference can work without a network and keep some processing local. It is constrained by memory, battery, hardware support and model size. Cloud processing can offer different resources but introduces connectivity, data-transfer and service considerations. A hybrid workflow can combine them. “On-device” describes where an operation runs; it does not prove that the entire application never uploads information. Check the whole data path.
Measure the experience that matters
Ask how long a result takes on the target device, what happens offline and which uncertain cases go to a person. Test with representative inputs instead of only convenient screenshots. The best model for a benchmark may not be the best fit for a field worker holding a phone in poor light. Preserve the original capture and the reviewer’s decision so later users can understand how a report was produced.
- Capture evidence
- Choose processing location
- Measure on target hardware
- Review uncertain results
Try it yourself
A field assistant must work with no signal. Name two questions you would ask before choosing a cloud-only design.
Compare with a worked answer
Can the essential task run locally, and what can the worker do when the service is unavailable? Also measure the target device’s memory, latency and battery behavior rather than assuming every phone performs alike.
Check your understanding
Choose one answer for each question, then check your score. Read the explanations and retry any question you missed.
Read the answer guide
1. Their visible appearance differs; capture conditions also need checking Lighting, angle and camera behavior can change appearance even without the environmental change you are investigating.
2. That a particular model operation runs on the device You still need to inspect the full application’s data flows and hardware constraints.
3. Representative inputs on the target devices, including offline behavior Deployment evidence should match the users, inputs and operating conditions of the intended workflow.
Was this explanation clear?
Help improve this lesson. With cookies accepted, HerbDev records one current rating per lesson for this browser. Server checks limit repeated submissions; accepted changes may also appear in site analytics. No written answer or account is needed.
You can also share a suggestion with HerbDev.
Review missed questions
Revisit questions missed on your latest checkpoint attempts. These short reviews help you practise; they do not change your best checkpoint scores. Return to the lesson checkpoint when you are ready.
Check a lesson’s answers to build your review list. Saved reviews are available after cookie acceptance.
Course complete. Keep the questions coming.
You have answered every checkpoint correctly. Use your worked exercises to explain the ideas to someone else, then apply them to a small project.
Apply it in the final project · Choose what to learn next →Keep exploring: What next? · AI vocabulary · ML 101 · Learning hub