Idanwo Aworan ifori

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Kini eto ifori aworan kan gbejade bi o ti njade?
   • Apejuwe ede adayeba ti aworan naa
   • Ẹya ti o ga julọ ti aworan naa
   • Akojọ awọn ipoidojuko piksẹli nikan
   • Faili fisinuirindigbindigbin ti aworan naa
2. Ninu opo gigun ti akole ti Ayebaye, ipa wo ni kooduopo wiwo ṣe?
   • O kọ ọrọ gbolohun ikẹhin nipasẹ ọrọ
   • O yọkuro awọn aṣoju ẹya lati aworan naa
   • O tumọ awọn akọle laarin awọn ede
   • O ṣe ikun girama ti akọle
3. Kini idi ti akiyesi wulo ni ifori aworan?
   • O rọ aworan naa lati ṣafipamọ ibi ipamọ
   • O yọ iwulo fun eyikeyi data ikẹkọ kuro
   • O jẹ ki awoṣe idojukọ lori awọn agbegbe aworan ti o yẹ fun ọrọ kọọkan
   • O ṣe iyipada akọle sinu ohun
4. Iru data wo ni o lo pupọ fun ikẹkọ ati igbelewọn ifori aworan?
   • MNIST
   • WikiText
   • LibriSpeech
   • MS COCO
5. Kini o tumọ si fun oluyipada akọle lati ṣe ipilẹṣẹ awọn ọrọ 'autoregressively'?
   • O ṣe asọtẹlẹ ọrọ kọọkan nipa lilo aworan ati awọn ọrọ ti o ti ipilẹṣẹ titi di isisiyi
   • O ṣe agbejade gbogbo awọn ọrọ nigbakanna
   • O ṣe agbejade ọrọ kan ṣoṣo fun aworan kan
   • O yi aworan pada si awọn piksẹli
6. Ipo ikuna ti o wọpọ ti awọn awoṣe ifori ti a mọ si 'hallucination' tọka si kini?
   • Ṣiṣejade awọn akọle ti o kuru ju
   • Apejuwe awọn nkan ti kii ṣe ni aworan gangan
   • Titumọ awọn akọle ti ko tọ
   • Nṣiṣẹ ju laiyara lori awọn aworan nla
7. How many human captions were collected for each image in the original COCO Captions dataset?
   • One caption total for the entire dataset
   • No text descriptions were collected
   • Five captions per image
   • One hundred captions for every image
8. What does CIDEr compare when evaluating an image caption?
   • Only the image’s pixel resolution
   • The caption’s audio waveform against a transcript
   • Whether the caption contains a certain number of words
   • The candidate caption against human reference captions using consensus-weighted n-grams