Idanwo ti o sopọ mọ itọsọna · Lile Ipele

Idanwo Ijinle Monocular

Assess monocular depth ambiguity, self-supervised view synthesis, relative scale, and version-specific depth-model evidence.

Jẹmọ itọsọna awọn ọnaIṣiro Ijinle Monocular
Ibeere 1 ti 8

Kini o jẹ ki iṣiro ijinle 'monocular'?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Kini o jẹ ki iṣiro ijinle 'monocular'?
    • O ṣe afihan ijinle lati aworan kan tabi kamẹra, kii ṣe sitẹrio tabi awọn sensọ
    • O nlo awọn kamẹra meji ti o ya sọtọ bi oju eniyan
    • O ṣiṣẹ nikan lori awọn aworan dudu-ati-funfun
    • O nilo ọlọjẹ lidar ti a so mọ kamẹra naa
  2. What does a monocular depth estimator predict for an image?
    • A stereo disparity field that requires two camera views
    • A per-pixel depth or relative-depth map from one camera view
    • A categorical object-label map with no distance estimate
    • A dense motion-vector field between two video frames
  3. What data recipe does the Depth Anything V2 paper report?
    • Train only on real labeled images and discard synthetic data
    • Train on unlabeled audio and convert speech to depth maps
    • Train a DINOv2-G teacher on synthetic labeled images, then use its pseudo-labels on real images to train students
    • Use stereo disparity as labels for every image at inference time
  4. Iwoye wo ni awọn nẹtiwọọki monocular KO gbẹkẹle taara fun ijinle?
    • Ojulumo iwọn ti awọn ohun
    • Sojurigindin gradients ati irisi
    • Occlusion (eyi ti ohun dina miiran)
    • Iyatọ sitẹrio laarin awọn kamẹra meji nigbakanna
  5. What does Monodepth2 use to reduce the effect of moving objects in self-supervised training?
    • An auto-masking procedure that excludes pixels inconsistent with camera motion
    • A learned text tokenizer
    • A region-proposal network
    • A fixed depth value for every object
  6. What new evaluation benchmark does the Depth Anything V2 paper introduce?
    • A benchmark of 2,000 full metric depth maps from LiDAR only
    • DA-2K, with sparse relative-depth pairs annotated across 1,000 images
    • A corpus of six public datasets and randomly captured photos
    • A camera-calibration set containing no depth comparisons
  7. How can self-supervised monocular depth training obtain a learning signal?
    • By reading depth labels from every pixel
    • By converting image colors directly to meters
    • By reconstructing one view from adjacent video frames and penalizing photometric mismatch
    • By using only language captions
  8. What does a relative-depth output indicate?
    • Exact meters for every pixel
    • Camera GPS position
    • Object identity only
    • Ordering or relative distance, not necessarily calibrated metric distance