Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Kutenganisha Chanzo cha Muziki wa Demucs

Kagua muundo wa wimbi la Demucs na utenganisho wa mseto, SDR kwenye MUSDB, na umakini wa kikoa cha HT-Demucs kwa data ya ziada ya mafunzo.

Njia za mwongozo zinazohusianaMgawanyo wa Chanzo cha Muziki wa Demucs
Swali 1 ya 8

Je, Demucs hufanya nini kwa wimbo uliokamilika?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Je, Demucs hufanya nini kwa wimbo uliokamilika?
    • Huitenganisha katika mashina kama vile sauti, ngoma, besi, na nyinginezo
    • Inabonyeza kuwa faili ndogo
    • Inatafsiri maneno
    • Huongeza kasi bila kubadilisha sauti
  2. Ni nini hufanya Demu Mseto kuwa 'mseto'?
    • Inachanganya nyimbo mbili tofauti pamoja
    • Inachakata sauti katika vikoa vyote viwili vya mawimbi na spectrogram
    • Inatumia maneno ya Kiingereza na Kifaransa
    • Inaendesha nusu kwenye CPU na nusu kwenye karatasi
  3. Which metric did the original Demucs paper report for separation quality on MUSDB?
    • Kiwango cha makosa ya Neno (WER)
    • Mean reciprocal rank (MRR)
    • Uwiano wa Mawimbi hadi Upotoshaji (SDR)
    • Makutano ya Muungano (IoU)
  4. Which research organization is associated with the original Demucs project?
    • The Allen Institute for Artificial Intelligence
    • Kyutai Labs
    • The International Society for Music Information Retrieval
    • Utafiti wa AI wa Facebook (FAIR)
  5. Kwa nini Demucs za mapema zilifanya kazi moja kwa moja kwenye muundo mbichi wa wimbi?
    • Ili kuepuka kupoteza taarifa za awamu
    • Kwa sababu spectrogram ni kinyume cha sheria
    • Kufanya faili kuwa kubwa kwa makusudi
    • Kwa sababu mawimbi hayana sauti
  6. Je, ni seti gani ya data iliyo muhimu kwa mafunzo na kuweka alama za Demucs?
    • Picha ya Picha
    • MUSDB18
    • MNIST
    • Utambazaji wa Kawaida
  7. What attention pattern does the HT-Demucs paper add between its waveform and spectrogram branches?
    • Attention only over lyric tokens
    • A single attention layer over stereo channels only
    • Self-attention within each domain and cross-attention across domains
    • No attention; only a nearest-neighbor lookup
  8. What did the HT-Demucs paper report about extra training songs?
    • Extra songs reduced SDR by 10 dB in every setting
    • The model reached the same result with zero training songs
    • The paper reported only a word-error-rate change
    • With 800 extra songs, HT-Demucs improved SDR by 0.45 dB over Hybrid Demucs trained on the same data