InoteveraGaidhi rinotevera
Adversarial Mienzaniso uye Kusimba
Tekinoroji
Gwara reVisual AI
ImageNet-C inoyedza kuti ma classifier emifananidzo anobata sei kukanganiswa kwakajairika kwakaiswa pamifananidzo yeImageNet validation, sekudengenyeka, ruzha uye maitiro anofanana nemamiriro ekunze.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
Inoyera rudzi rwekuchinja kwekuparadzirwa uko kuverenga kwechokwadi kwemifananidzo kunogona kusakanganisa. Seti yayo yekuvhiringidzwa kwekunyepa inoshanda pakuedza asi haina kusimbisa kusimba kune kamera, nharaunda kana kurwisa kwechokwadi kwese.
An image classifier may perform well on clean validation photos and fail when the same objects are blurred, noisy or altered by weather-like effects. ImageNet-C, introduced by Hendrycks and Dietterich, provides standardized corruptions and severities applied to ImageNet validation images. Because the underlying image labels remain the same, it isolates a family of input changes for comparative testing. It is not simply a second training set, and its intended purpose differs from adversarial examples crafted to fool a particular model. Corruption categories include forms of noise, blur, weather and digital changes. A model can be strong on one family and weak on another. A mean score is useful for ranking under a specified protocol, but per-corruption and per-severity results show where the failures occur. Clean accuracy and corruption robustness can differ; a better clean score does not automatically mean better behavior under every disturbance. The benchmark has standard images and transformations, which helps reproducibility across methods. The protocol has limits. Synthetic snow on a photo is not the same as a real camera used during a storm; compression, sensor processing and exposure interact with conditions. ImageNet-C also tests a particular category set and label ontology. A warehouse inspection model or road camera needs tests from its own devices and failure modes. Avoid using test labels repeatedly to tune a model and then presenting the same benchmark as independent evidence. Keep a held-out local evaluation and document training augmentations. Robustness evaluation should connect errors to decisions. A minor misclassification in a photo organizer differs from a missed hazard in a robotics task. Compare error rates and confidence under realistic corruptions, examine whether the model can recognize degraded input, and define a fallback when evidence is poor. ImageNet-C is a valuable stress test, not a complete safety certificate.
Chikwata chinovheneka mhedzisiro yakachena yeImageNet ye classifier pamwe nekukanganisa pane dzakasiyana-siyana dzekudengenyeka kweImageNet-C.
Chigadzirwa chemufoni wepamhino chinoyedza kusviba kwepasi uye kudengenyeka kwekufamba kubva kune zvigadzirwa zvake pachake pamwe chete ne benchmark yenyika yose.
Mutsvakurudzi anoratidza mhedzisiro yega yega yekukanganiswa panzvimbo pekuvanzisa kukundikana kukuru mumhedzisiro yakabatana.
Muenzi wechengetedzo anopatsanura kuita kwekuita pasi pekurira kwemifananidzo kunotaridzika sekwakasikwa kubva pakukanganiswa kweadversarial kunoitwa nechinangwa.
Maitiro ekushandisa zviyero zvekuvhiringidzwa acharamba achishanda pakuedza zvakarongeka, uye shanduro itsva dzinogona kusanganisira maitiro echokwadi e kamera nemamiriro ekunze. Hapana runyorwa rusingaperi runogona kukwanisa kusanganisira mamiriro ose. Mapoka anofanira kusanganisa bvunzo dzekushushikana dzeveruzhinji nemashoko kubva kune zvigadzirwa zvavo uye kuongorora mashandiro sezvo zvigadzirwa izvi zvichichinja. Modhi dziri nani dzinogona kuvandudza avhareji ichichengetedza nzvimbo dzisina kuona kune imwevhiringidzwa kana kukomba, saka mishumo yakadzama inokosha. Zvigadzirwa zvinogonawo kuona mhando yakaipa yezvinhu uye kukumbira mufananidzo wechipiri kana kudzikisa kusarudza. Kuvimbika kunofanira kutsanangurwa nezve basa remushandisi uye mutengo wekukundikana, kwete nhamba imwechete ye benchmark.
Tsamba yemahara
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
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
ImageNet-C inoyedza kuti ma classifier emifananidzo anobata sei vhiringidzwa rakajairika rakapihwa pamifananidzo ye validation ye ImageNet, senge kusvibiswa, ruzha uye maitiro anofanana nemamiriro ekunze. Inoyera rudzi rwekuchinja kwekuparadzirwa kunyangwe kuverenga kwemifananidzo yakachena kunogona kusakanganisa. Seti yayo ye synthetic corruption inoshanda pakuedza asi haina kusimbisa kuvimbika kune kamera chaiyo, nharaunda kana kurwisa kwekushungurudza.
Kuchinja kwakarongeka kunochinja pikicha asi chikamu chinangwa chinoramba chakafanana.
ImageNet-C inoshanda nezvikanganiso zvakagadzirirwa panzvimbo pekushungurudza kwakatarisana ne model.
Kushandurwa kunowanikwa neruzhinji kunongofungidzira zvimwe chete zvemaitiro echokwadi ekutora mifananidzo.
Mhinduro inodzokororwa inogona kushandura benchmark yakasiyiwa kuita data rekusimudzira.
Mamiriro ekunze ekuziva chaiwo anoda humbowo hwekusiyana.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Adversarial Mienzaniso uye Kusimba
Tekinoroji