Bài kiểm tra có liên kết hướng dẫn · Cứng Cấp độ

Câu hỏi về Mạng chiếm chỗ và Hình dạng 3D ngầm

Kiểm tra các truy vấn chiếm chỗ, nhãn bên trong-bên ngoài, trích xuất bề mặt tiềm ẩn, sự mơ hồ trong một chế độ xem, độ phân giải truy vấn và giới hạn hình dạng đào tạo.

Đường dẫn hướng dẫn liên quanMạng chiếm dụng 3D tiềm ẩn
Câu hỏi 1 của 8

What does an occupancy network predict when given a 3D query point?

Xem trước mọi câu hỏi8 câu hỏi

Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.

  1. What does an occupancy network predict when given a 3D query point?
    • The exact signed distance to the nearest surface
    • Whether that point is inside or outside a learned shape
    • The surface normal at a known mesh vertex
    • The RGB value of a camera pixel
  2. How is a mesh commonly obtained from an occupancy function?
    • Read a fixed triangle list directly from one image pixel
    • Treat the original image grid as a 3D surface
    • Sample the function in space and extract a threshold boundary
    • Connect only observed camera pixels without 3D queries
  3. Why is the hidden back of a single-view reconstruction uncertain?
    • The RGB view directly measures its hidden backside
    • A finer extraction grid creates new observed data
    • A high occupancy score certifies hidden geometry
    • The network relies on learned shape patterns where the image gives no direct observation
  4. If query resolution is increased after training, what improves and what does not automatically improve?
    • Mesh sampling may get finer, but missing learned shape detail is not recovered
    • The training data automatically expand to new categories
    • The conditioning view acquires an unseen backside
    • The learned function gains detail it did not encode
  5. Which training signal supports an occupancy network’s inside-outside task?
    • Only 2D object names without spatial labels
    • Labels for sampled 3D points relative to known shape surfaces
    • Only the model’s output mesh file size
    • Only camera timestamps without shape examples
  6. A model trained mainly on chairs is applied to an unusual sculpture. What risk grows?
    • The training prior necessarily covers every new category
    • A finer query grid removes domain mismatch
    • The coordinates stop representing 3D locations
    • Its learned shape prior may not generalize to unfamiliar geometry
  7. Why is an implicit occupancy shape different from a fixed voxel grid?
    • The function can be queried at chosen coordinates instead of storing only preset cell values
    • It always measures the real surface directly
    • It stores only preselected cell values like a voxel array
    • It makes finite mesh-extraction resolution unnecessary
  8. Before using an inferred hidden surface for fabrication, which check is strongest?
    • Accept a high occupancy score as physical measurement
    • Increase mesh gloss to conceal uncertain areas
    • Compare it with independent measurements or additional views
    • Use a finer query grid without new observations