Culture and technology guide
How HandFuture works
Follow the data from choosing a photo through hand-joint detection and reflection-card selection, including what remains in the current browser session and what the tool deliberately does not infer.
Publisher: HandFuture ยท Updated:
Choose a photo
The uploader accepts JPEG, PNG, or WebP files up to 10MB. Choose a photo that contains one hand. Browser FileReader reads the file you explicitly select into a temporary data URL and places it in the current page's application state.
There is no HandFuture upload endpoint for this photo. The photo is not uploaded to HandFuture, and it does not become account data or public content; file selection only makes the image available to the current browser page for detection. You should still keep faces, documents, and identifying surroundings out of the frame.
Locate hand joints
MediaPipe looks for hands and returns standard hand landmarks plus handedness information. HandFuture first confirms that only one hand was found, then validates that one result contains exactly 21 finite joint coordinates before treating the detection as usable.
The interface keeps the original uploaded photo and reports the detection status in text; after validation succeeds, the reflection-card button becomes available. The landmark model does not identify the life, head, heart, or fate creases, and it supplies no conclusion about the person in the photo.
Select a reflection card
After successful detection, and only after you press the button, the program turns normalized coordinates into a fixed geometry signature. That signature selects one of four general prompts. Identical normalized coordinates return the same key, so the result is not a fresh random draw on every click.
This step makes no inference about personality, compatibility, health, career, wealth, or future events. A card is an open-ended question from a fixed set for non-scientific cultural entertainment and personal reflection. Geometry provides a stable selection key; it is not a measurement of a person.
Clear and understand limits
Reset clears the photo, detection result, and reflection card from application state. Closing or refreshing the page clears the photo from memory. Disclaimer acceptance alone persists in local storage; neither the photo nor landmark coordinates persist with that preference.
Detection can fail because of incomplete framing, poor light, unsupported image decoding, or network and content blocking while the model loads. No-hand, multiple-hand, and unavailable-model results are limitation states. Changing the photo may help, but no detector result amounts to an interpretation of palm creases.