InoteveraGaidhi rinotevera
Siamese Networks uye Triplet Kurasikirwa
Tekinoroji
Nhungamiro yehunyanzvi
Focal Loss for Inbalanced Detection
Focal kurasikirwa ibasa rakagadziridzwa kurasikirwa iro pasi-uremu mienzaniso iri nyore saka detector inogona kutarisa kune yakaoma, isingawanzo.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
Yakagadzirisa iyo yakanyanyisa kumashure-kupesana-chinhu kusaenzana iyo yakaremara imwe-nhanho chinhu detectors.
Kudzika Kwakadzika
Mukuona kwechinhu, mufananidzo unogona kunge uine zvinhu zvishoma chaizvo asi makumi ezviuru enzvimbo, dzinenge dzese dziri nyore kumashure. Neyakajairwa muchinjiko-entropy, aya mafashama ezvakareruka zvakashata anotonga gradient uye anonyura kunze zvisingawanzo hunhu. Focal kurasikirwa, yakaunzwa mu2017 RetinaNet bepa naLin nevamwe vaanoshanda navo paFacebook AI, inogadzirisa izvi nekuwanza muchinjiko-entropy nechinhu (1 - p_t) ^ gamma. Kana sample yaiswa pazvitemba uye nemazvo, p_t iri pedyo ne1, saka chinhu chinodzikira kuenda ku zero uye muenzaniso wakasarudzika haubatsire. Mienzaniso yakaoma, yakasarudzika inochengeta huremu hwakazara. Iine gamma yakatenderedza 2, RetinaNet inofananidzwa kana kurova inononoka-nhanho mbiri-nhanho seFaster R-CNN ichiramba iri nyore-inopfuura network.
Kuitwa muNyika yeChokwadi
Kuona zvikwangwani zvidiki zvemumugwagwa kana vanofamba netsoka vari kure mumafuremu ekutyaira uko mapixel mazhinji ari kumashure.
Kuwana mamota asingawanzo kana maronda mumakiriniki ekurapa anotongwa nenyama ine hutano.
Kuona hurema pamutsetse wekugadzira uko ruzhinji rwezvikamu zvakaongororwa zvakajairika.
Kuziva zvikepe zvidiki kana mota mune hombe satellite uye mifananidzo yemuchadenga.
Ramangwana reKurasika Kwekutarisisa Kwekuonekwa Kusina Kusimba
Kurasika kwekutarisa kwave chinhu chakasarudzika kupfuura RetinaNet, inoonekwa mumadhijitari akaita seFCOS, muzvikamu, uye mune refu-muswe classification. Misiyano yakadai semhando yekurasika kwemhando, kugovera kwakatarisa kurasikirwa, uye kurasikirwa kwevarifocal kunonatsiridza kune yemazuva ano anchor-isina uye transformer-based detectors. Sezvo kucherechedzwa kuchichinja kuenda ku-set-prediction modhi seDETR inoshandisa bipartite matching, focal-style reweighting inoramba iri chishandiso chinoshanda pese painotsveyamiswa makirasi.
Kuzvishandisa
Nei zvichikosha
- Mutengo uye bhajeti. Zvisarudzo zvezvivakwa zvinotyaira kuita uye mutengo wekushandisa kwemakore.
- Sarudzo dzakajeka. Dzidzo yehunyanzvi inobatsira zvikwata kusarudza murwi wakakodzera, kwete iwo mutsva chete.
- Kudzora kwemhando yepamusoro. Sarudzo dzeinjiniya dziri nani dzinoderedza zviitiko zvekuvimbika mukugadzira.
Chenjera
- Kugadzirisa imwe bhenji kunogona kuvanza yakafara system kushaya simba.
- Infrastructure uye mari yekugadzirisa inowanzotarisirwa pasi.
- Chengetedzo uye kucherechedzwa mapundu anogona kukura sezvo masisitimu anowedzera kuoma.
Matanho anotevera
- Tsanangura latency, mhando, uye mutengo zvinangwa usati waitwa.
- Benchmark pasi pechokwadi mutoro uye data mamiriro.
- Chishandiso chekutarisa zvikanganiso, kudonha, uye mushandisi maitiro.
- Gadzirira nzira dzekudzosera kumashure uye dzezviitiko usati wawedzera.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the Focal Loss for Imbalanced Detection quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
Chii chinonzi Focal Loss for Inbalanced Detection?
Focal kurasikirwa ibasa rakagadziridzwa kurasikirwa iro pasi-uremu mienzaniso iri nyore saka detector inogona kutarisa kune yakaoma, isingawanzo. Yakagadzirisa kunyanyisa kwemashure-kuchipikisa-chinhu kusaenzana kwakaremadza chimwe-nhanho chinhu chinoona.
Nderipi dambudziko guru muchikamu chimwe chekuonekwa kwechinhu chaive chekurasika kwakagadzirirwa kugadzirisa?
Mukuona kwakaomarara, mienzaniso iri nyore yekumashure inodarika zvinhu chaizvo uye inotonga yakajairwa muchinjiko-entropy gradient. Focal kurasikirwa reweights saka isingawanzo, yakaoma mienzaniso yakawanda.
Ndeipi iyo modulating factor iyo focal kurasikirwa inowanda pacross-entropy?
Kurasika kwakananga ndiko -(1 - p_t)^gamma * log(p_t). Iyo (1 - p_t)^izwi regamma rinodzikira kuenda ku zero kuitira fungidziro dzechokwadi.
Chii chinoitika pakurasikirwa kwekutarisa kana iyo inotarisisa parameter gamma yakaiswa ku0?
Pa gamma = 0 chinhu (1 - p_t) ^ 0 yakaenzana ne1, saka kurasikirwa kwepakati kunoderedza chaizvo kusvika kune yakajairwa muchinjiko-entropy.
Chii chinoita kuwedzera gamma kune mienzaniso iri nyore, yakanyatsorongedzwa?
Yepamusoro gamma inoita iyo (1 - p_t) ^ izwi regamma ridzike nekukurumidza kune yakakwira p_t, saka mienzaniso iri nyore inopa kunyange kushoma pakurasikirwa uye gradient.
Ndeipi detector architecture yakaunza focal kurasikirwa?
Focal kurasikirwa kwakaunzwa mu2017 RetinaNet bepa, ichigonesa danho-rimwe detector kukwikwidza kwakaringana-matanho maviri nzira.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi