Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Apa Ohunkohun Awoṣe adanwo

Understand promptable image and video segmentation through the original SAM and its later SAM 2 extension, data and memory methods.

Jẹmọ itọsọna awọn ọnaApa Ohunkohun Awoṣe
Ibeere 1 ti 8

Ero pataki wo ni o jẹ ki SAM jẹ 'awoṣe ipilẹ' fun ipin?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Ero pataki wo ni o jẹ ki SAM jẹ 'awoṣe ipilẹ' fun ipin?
    • O ṣe itọju ipin bi iṣẹ-ṣiṣe ti o yara ti o ṣe akopọ si awọn nkan ti a ko rii
    • O ṣiṣẹ nikan lori awọn oju
    • O nilo isọdọtun fun gbogbo aworan tuntun
    • O ṣejade awọn apoti didi, kii ṣe awọn iboju iparada
  2. Ni aijọju bawo ni o ṣe tobi dataset SA-1B ti a lo lati ṣe ikẹkọ SAM?
    • Nipa awọn iboju iparada 1,000
    • Ju awọn iboju iparada 1 bilionu kọja awọn aworan miliọnu 11
    • Nipa awọn iboju iparada 1 milionu
    • Gangan 50,000 awọn iboju iparada
  3. Kini idi ti SAM ṣe pin faaji rẹ si koodu koodu ti o wuwo ati koodu aiyipada kiakia kan?
    • Lati jẹ ki ikẹkọ ko ṣeeṣe
    • Lati yago fun lilo eyikeyi netiwọki nkankikan
    • Nitorinaa ifibọ aworan kan le tun lo fun ọpọlọpọ awọn itara ibaraenisepo ni akoko gidi
    • Nitori awọn iboju iparada ti wa ni iṣiro lori Sipiyu nikan
  4. Bawo ni SAM ṣe mu itọsi aibikita kan, bii titẹ ẹyọkan ti o le tumọ si awọn nkan pupọ?
    • O kọ lati jade ohunkohun
    • Nigbagbogbo o mu nkan ti o tobi julọ
    • O beere lọwọ olumulo lati tẹ apejuwe kan
    • O ṣe asọtẹlẹ awọn iboju iparada pupọ ati ṣe ipo wọn nipasẹ igbẹkẹle
  5. Iru egungun ẹhin wo ni SAM nlo lati fi koodu sii aworan titẹ sii?
    • Ayipada Iran kan (ViT)
    • Nẹtiwọọki ti nwaye loorekoore
    • Igi ipinnu
    • A atilẹyin fekito ẹrọ
  6. Agbara pataki wo ni SAM 2 ṣafikun ni ọdun 2024?
    • Igbasilẹ ohun
    • Pipin ati ipasẹ awọn nkan kọja awọn fireemu fidio
    • Ṣiṣẹda awọn akọle ọrọ
    • 3D titẹ sita support
  7. What role did the data engine play in the original SAM project?
    • It replaced the image encoder with a text-only model
    • It automatically verified every mask as medically correct
    • It iteratively combined model-assisted annotation and human annotation to build SA-1B
    • It generated labels without human or model assistance
  8. How does SAM 2 use memory for video segmentation?
    • It discards all frames after processing each one
    • It predicts only a video-level class label
    • It replaces the prompt with an audio transcript
    • It stores information from prior frames to help segment the object in later frames