Kuiz berkaitan panduan · Sederhana Tahap
Penyampaian Teks dalam Kuiz Imej Dijana AI
Menguji sebab token subkata CLIP menyembunyikan ejaan, cara pengekod peringkat T5 dan bait mempertingkatkan huruf, dan cara latihan menyedari glif dan bantuan lukisan.
soalan 1 daripada 8
Mengapakah model imej berasaskan CLIP bergelut untuk mengeja perkataan dengan betul?
Pratonton setiap soalan8 soalan
Jawapan dan penjelasan muncul semasa anda mengambil kuiz.
- Mengapakah model imej berasaskan CLIP bergelut untuk mengeja perkataan dengan betul?
- CLIP had no vocabulary at all
- CLIP turns prompts into subword tokens that capture meaning, not letter sequences
- CLIP processes only numbers
- CLIP was trained only on images that contained no text
- Pengekod jenis yang manakah yang didapati lebih baik oleh kajian sedar watak Google?
- A smaller CLIP encoder
- An audio encoder
- An image-only encoder
- A byte-level encoder such as ByT5 that sees individual characters
- Mengapakah ralat kecil boleh diterima dalam muka yang dijana tetapi tidak dalam teks yang dijana?
- A single wrong stroke can change which letter it is
- Faces are generated by a separate model
- Text is always drawn at a higher resolution
- Faces use a different kind of noise
- Apakah had dalam data latihan yang menyebabkan sukar untuk memautkan perkataan dengan bentuk huruf?
- Training images used too many fonts
- All training images were black and white
- Captions rarely wrote out the text visible in the image
- Only English-language images were included
- Apakah yang ditambahkan oleh sistem penyelidikan seperti TextDiffuser dan GlyphControl pada penjanaan imej?
- Invisible watermarks
- Explicit layouts or pre-rendered letter images to guide the model
- Audio narration
- Faster sampling and nothing else
- Which model does the guide say paired its image generator with a T5 encoder?
- Stable Diffusion 3
- Resapan Stabil 1.5
- Stable Diffusion 1.4
- The original CLIP model
- For production work that needs exact text, what does the guide call the most reliable approach?
- Writing a much longer prompt
- Menurunkan resolusi
- Adding a heavy negative prompt
- Adding text outside the model, or inpainting a masked area with the corrected word
- According to the guide, how do modern image models produce letters?
- By copying shapes from a font file
- By running OCR on the prompt
- By generating them as pixels, which is why long passages and small print still go wrong
- By looking up images of words on the web