SkinScreen

SkinScreen, LLC.

Mayelana nalolu hlelo lokusebenza

Uhlelo lokusebenza lwe-SkinScreen lwandisa amandla womuntu ekutholeni nasekuhlukanisweni kwezilonda zesikhumba / umdlavuza wesikhumba kuzinhloso zokusekela ezempilo ezisuselwa enanini lokuxhaswa. I-SkinScreen inikeza amandla okuthola izilonda zesikhumba ezinobungozi futhi ezinobungozi ngesikhathi sangempela ngesisombululo esinembe kakhulu nesinembile. Isixazululo sisebenzisa amandla okufunda okujulile, indlela engaphansi kobuhlakani bokufakelwa (i-AI), ukuvumela ukubikezela okusheshayo nokunembe kakhulu kunangaphambili. Ngegama esilithengisile, elibizwa nge-Indescribable Model, liyimodeli ye-AI ehlwanyelwe ama-hyperparameter ekuqaleni kodwa imodeli iyaqhubeka nokuziqeqesha ekutholeni okulingana kakhulu nedathasethi ngaphandle kokungenelela komuntu kwesikhathi esizayo okudingekayo. Njengamanje, ukutholwa kwenziwa ngesandla ngudokotela wesikhumba noma uchwepheshe ngendlela ye-heuristic eyaziwa njenge-ABCDE (Asymmetry, Border Irregularity, Colour, Diameter, Evolution).

ISkinScreen inikeza umehluko omningi kunezinye izixazululo emakethe:

1. Qinisekisa ubumfihlo bomsebenzisi - Ngokusebenzisa ukwakhiwa kwakamuva kwe-MobileNetV2 imodeli ye-AI iyakwazi ukusebenza kudivayisi yomsebenzisi futhi azikho izithombe ezidinga ukulayishwa zibuyele emuva kumaseva we-SkinScreen ngokungafani nezinye izisombululo.

2. Thola ukuthi isilonda sesikhumba sikhona yini - Izixazululo eziningi zokutholwa kwesikhumba ze-AI azibonisi ukuthi isilonda sesikhumba sikhona yini esithombeni ekuqaleni. Bathembele ekungeneleleni okwenziwe ngesandla ngumsebenzisi womuntu ukuhlinzeka isithombe sesilonda sesikhumba. Isibonelo, uma umsebenzisi enikeza isithombe sendlulamithi izixazululo zabo zizohlukanisa isithombe kungakhathalekile. Imodeli ye-AI eyinkimbinkimbi yeSkinScreen iyakwazi ukubona ukuthi ngabe isilonda sesikhumba sikhona ngaphambi kokuhlukaniswa.

3. Thola izigaba eziningi zezilonda zesikhumba - Ngokuthola amakilasi avamile anobungozi futhi abulalayo ezilonda zesikhumba (Actinic Keratoses, Angioma, Basal Cell Carcinoma, Dermatofibroma, Melanocytic nevus, Melanoma, Seborrheic keratoses, Squamous Cell Carcinoma, Vascular lesions) siyakwazi nikeza ngempendulo engcono kumuntu ngamunye oxhuma ne-SkinScreen. Futhi siyaqhubeka nokwandisa inani lamakilasi ezilonda zesikhumba esiwasekelayo.

4. Nikeza ngokunemba okuphezulu kanye namanani okunemba - Sisebenzisa indlela ekabili yokufeza amazinga aphezulu okunemba nokunemba. Siqale sisebenzise ukwahlukaniswa kwesigaba esisodwa ukukhomba ukuthi isilonda sesikhumba sikhona yini esithombeni. Uma kunjalo, khona-ke siyakwazi ukubuyisela emuva amakilasi ama-3 okungenzeka abe nezilonda zesikhumba kanye namathuba ahlobene nawo. Ingxenye yalokhu ifezwa ngezithombe eziyi-180,000 esizisebenzisayo ukuqeqesha imodeli yethu ye-AI.

5. Nikeza impendulo yesikhathi sangempela - i-SkinScreen iyakwazi ukuhlinzeka ngemiphumela emuva kumsebenzisi ngaphansi kwemizuzwana emibili ngokwesilinganiso. Ngokusebenzisa ukwakheka kwe-MobileNetV2 okune-latency ephansi nokunemba okuphezulu kanye nezithuthukisi ezimbalwa zokuphathelene siyakwazi ukwazisa umsebenzisi ngemiphumela ngesikhathi esifanele.

6. Nikeza ngamathuluzi asebenziseka kalula - Amapulatifomu ahlukene e-SkinScreen akwazi ukusiza abasebenzisi ekuxhumaneni kwabo nethuluzi. Sizama ukufeza lokhu ngokusebenzisa amathuluzi okusekela abalulekile ekutholeni izilonda zesikhumba kungakhathalekile isizinda somsebenzisi namakhono akhe.

Yini entsha

- Updated to align with latest SDK (v36)