InoteveraGaidhi rinotevera
Depthwise Separable Convolutions
Tekinoroji
Gwara reVisual AI
Deformable Convolutions
Deformable convolutions rega neural network ikotame gidhi yesampling kuti iteedzere chimiro chaicho chezvinhu pachinzvimbo chekumanikidza kuburikidza neyakaomarara sikweya hwindo.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
This makes models far better at handling odd shapes, scale changes, and geometric distortion.
Kudzika Kwakadzika
Yakajairika convolution sampuli mapixels payakagadziriswa offset - yakarongeka 3x3 grid yakatarisana nenzvimbo yega yega. Izvo zvinoshanda zvakanaka kune maumbirwo asi zvinonetsa kana zvinhu zvakarereka, zvakatambanudzwa, kana kuumbwa zvisingaite. Deformable convolutions, yakaunzwa naDai nevamwe vaaishanda navo ku Microsoft Tsvagiridzo muna 2017, wedzera diki diki rakadzidzwa kune imwe neimwe yemasampling poindi. Iyo network inotarisa kupinza uye inofanotaura shanduko ye 2D kune yega yega gidhi chinzvimbo, saka iyo inogamuchira inokwanisa kutenderera kuti imbundire mupendero wakakomberedzwa kana kutevera nhengo yakatenuka. Deformable RoI pooling inoshandisa iyo imwe pfungwa kune dunhu maficha. Version 2 (2018) yakawedzera per-point modulation uremu, ichirega iyo layer ichinyorovesa kana kukwidziridza sampuli yega yega, iyo yakarodza chinhu-yekuona chokwadi pamabhenji akaita seCOCO.
Kuitwa muNyika yeChokwadi
Kuonekwa kwechinhu paCOCO, uko mativi akaremara anowedzera huchokwadi pazvinhu zvakarebesa kana zvakatenderedzwa sezvitima netwiza.
Semantic segmentation yezviratidziro zvemumugwagwa, kubatsira mamodheru kuteedzera mitsetse yakakomberedzwa uye zvimiro zvezvivakwa zvisina kujairika.
Deformable DETR yekuona-kumagumo-kusvika-kumagumo, uchishandisa yakadzidza offset kuita kuti transformer itarise ishande
Kufungidzira kwekurapa, uko mapundu uye nhengo dzine maumbirwo asina kuomarara ayo anogadzika grids anobatwa zvisina kunaka
Ramangwana Rekuremara Convolutions
Kuremara kutarisa kwave musana wekuonekwa kwemazuva ano: Deformable DETR inoshandisa yakadzidzwa sampling offsets kuita kuti transformer itarise kushomeka uye nekukurumidza, kucheka nguva yekudzidziswa zvakanyanya maringe neiyo yekutanga DETR. Tarisira musimboti wakagomara kuti urambe uchipararira kuita vhidhiyo, makore e3D, uye mhando dzemutauro wekuona, uko sampling inogadziriswa inobatsira kubata kufamba, kuvharika, uye geometry isina kujairika. Sezvo tsigiro yehardware yekusajaira ndangariro yekuwana inovandudza, vanobatika vanobatika vanofanirawo kuwana zvakachipa uye zvakanyanya kuisirwa pamipendero michina.
Kuzvishandisa
Nei zvichikosha
- Kumhanya uye chiyero. Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero.
- Vaka sarudzo. Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma.
- Team uye workflow. Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa.
Chenjera
- Kodzero dzemifananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana provenance isina kujeka.
- Kuita kwemuenzaniso kunogona kusiyanisa kupenya, huwandu hwevanhu, uye nharaunda.
- Manyepo enhema anogona kusacherechedzwa kunze kwekunge zvikumbaridzo zvekuvimba zvikatariswa.
Matanho anotevera
- Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa.
- Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira.
- Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura.
- Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset.
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 Deformable Convolutions 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
What is Deformable Convolutions?
Deformable convolutions rega neural network ikotame gidhi yesampling kuti iteedzere chimiro chaicho chezvinhu pachinzvimbo chekumanikidza kuburikidza neyakaomarara sikweya hwindo. Izvi zvinoita kuti mamodheru ave nani pakubata zvimiro zvisinganzwisisike, shanduko yechikero, uye kukanganisa kwejometri.
Ko convolution yakagomara inowedzera chii kana ichienzaniswa neyakajairwa convolution?
Deformable convolutions inofanotaura yakadzidzwa offset (dx, dy) kune yega sampling nzvimbo, ichirega iyo grid kutenderera kuti ienderane nechinhu maumbirwo.
Ko sampling offsets mune deformable convolution inogadzirwa sei?
A parallel convolution bazi rinoburitsa maoffsets, ayo anodzidzwa kupera-kusvika-kumagumo kuburikidza nebackpropagation.
Nekuti zvakafanotaurwa zvigadziriso zvinowanzoita zvidimbu, komisheni yepixel inotorwa sei munzvimbo idzodzo?
Fractional offsets inoda kududzirwa kwebilinear, izvo zvinoita kuti oparesheni irambe ichisiyaniswa pakudzidziswa.
Chii chakawedzera Deformable ConvNets v2 (2018) pamusoro pepakutanga?
v2 yakaunza modulation, huremu hwakadzidzwa pane imwe sampling point iyo inogona kukwidziridza kana kudzvanya simba rayo, kuvandudza huroyi.
Sei mabhindauko akaremara achinyanya kubatsira pazvinhu zvisina kuumbwa zvakanaka?
Nekupeta gidhi resampling, munda unoshanda wekugamuchira unoenderana neiyo chinhu geometry pachinzvimbo chesikweya yakaoma.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi