Loading the AI model (about 14 MB, one download)...
?processor cores this device reports
...to download and prepare the model
MobileNetV2the model: 3.4 million learned numbers, 1,000 things it can name
1. Pick a test image
Click an image to classify it once. Each click is one run, and the time appears below and in your results table.
2. The answer
No result yet. Pick an image above.
The AI thinks this is:
Thinking time on this device:
What is being timed? Only the thinking: the moment the prepared picture goes into the model until its answer comes out. Downloading the page and the model is not counted, because that only happens once.
3. Your results
Copy these into your timing sheet. You need three runs of each image on each device. Then take the median: sort the three times and pick the middle one.
| Image | Run | Answer | Confidence | Time (ms) |
|---|---|---|---|---|
| Nothing yet. | ||||
1,000 milliseconds (ms) = 1 second.