Mashandiro ekubvisa kumashure kumashure kunoshanda
Iyi ndiyo yekugezera-dhayagiramu shanduro - zvakakwana kuvimba pombi, kwete pepa.
1. Input inova mapixels
Iwe unopa faira kana bitmap (donhedza, picker, paste, kana peji rekutora). Mupepeti anoiisa kune mufananidzo bitmap. Mafomati seHEIC anoda imwe nhanho yekuwedzera decode; kutadza kunomira nekukasira nekanganiso pachinzvimbo che canvas isina chinhu.
2. Kugadzirira kwemuenzaniso
Inference inoshandisa pane-mudziyo segmentation modhi (ISNet-kirasi ONNX) inomhanya neWeb Worker uye browser ONNX yekumhanyisa nguva. Kana uremu husati hwavharwa, unoona kurodha splash. Caching inodzivirira kudzokorora iyo hit nguva dzese.
3. Segmentation → mask
Mushandi anodzosera alpha mask (chii chiri pasi vs kumashure). Iyo UI inosanganisa chinyorwa pamusoro pechekibhodhi kana pamusoro pekumashure kwawakasarudza (ruvara, gradient, mufananidzo).
4. Kunatsa kwevanhu
Wedzera / dzima mabhurashi gadzirisa mask inoshanda pane CPU canvas. Kupfava kunochinja kuti sitiroko yakaoma sei. Export inoshandisa mask yaunoona - kwete kukanganwa otomatiki-chete snapshot.
5. Nhoroondo uye ekisipoti
Iyo pfupi IndexedDB nhoroondo inogona kuchengeta ichangoburwa yekutanga / mhedzisiro pairi. Dhawunirodha uye Copy emit PNG ine pachena yazvino kana kuzadza.
Manotsi ekuita iwe unozoona
- Kutanga kumhanya: network + init mutengo weiyo modhi. -Mifananidzo mikuru: inogona kugadzirisa pakuderedzwa kwemukati kugadzirisa neyambiro; kunze kuchiri kutarisisa saizi yakazara. -UI tambo inogara ichishandiswa nekuti inorema inference iri mumushandi - kana mushandi akadonha, unowana kukanganisa kwevashandi, kwete tebhu yakaomeswa zvachose.
Zvekuvanzika zve “nzvimbo” zvinoiswa paChakavanzika. Kudzvanya kwezuva nezuva: Tanga.