Prueba de iluminación de imágenes con IA

Las respuestas y explicaciones aparecen a medida que realiza el cuestionario.

1. What does image relighting aim to change?
   • The lighting, while keeping the subject's identity and shape
   • The subject's identity
   • The camera lens
   • El formato del archivo de imagen
2. What does intrinsic image decomposition separate?
   • Foreground and background
   • Red, green and blue channels
   • Albedo and shading
   • Text and images
3. What are light stages used for in portrait relighting research?
   • Rendering fonts
   • Comprimir vídeo
   • Training language models
   • Recording a person under many controlled lighting conditions
4. What physical principle does IC-Light's training enforce?
   • Conservation of momentum
   • An object's appearance under mixed lights equals the mix of its appearances under each light
   • Light travels faster in glass
   • Shadows are always black
5. Why is a color filter not true relighting?
   • It cannot move shadows or add highlights on the side facing a new light
   • It changes the subject's identity
   • It requires a GPU
   • Solo funciona en video
6. How is the foreground image given to IC-Light's denoiser?
   • As a text prompt
   • As a CLIP score
   • Through a separate audio encoder
   • As latents concatenated with the noisy latent as extra channels
7. Which practical step helps avoid glowing edges when relighting a cut-out subject?
   • Raising the step count to 500
   • Converting to grayscale
   • Making a clean matte of the subject first
   • Removing the background prompt
8. Why does better relighting matter for image authenticity?
   • It makes metadata impossible
   • Inconsistent shadows are a clue to composites, so better relighting makes fakes harder to spot
   • It always adds watermarks
   • It prevents image editing