Datasets:
File size: 4,285 Bytes
6c1c833 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | {
"field_id": "F004",
"field": "3d reconstruction",
"directions": [
{
"direction_id": "G0",
"name": "Scene and object recovery from visual observations",
"scope": "Recover 3D geometry, surfaces, and renderable representations of scenes or objects from images, RGB-D observations, drawings, or event-camera observations. This includes per-scene optimization, transferable inference, generative-prior completion of unseen regions, direct prediction from a single depiction, multi-view depth fusion, and event-measurement modeling; individual approaches need not combine these conditions."
},
{
"direction_id": "G1",
"name": "Human and hand–object reconstruction",
"scope": "Recover the shape and appearance of bodies, heads, faces, or hands, including personalized representations controllable by pose or expression. This also includes using hand–object relationships to recover a manipulated object or both participants in an interaction. Not every approach must produce an animatable person or reconstruct both hand and object."
},
{
"direction_id": "G2",
"name": "Dynamic and articulated reconstruction",
"scope": "Recover 3D shape and motion that vary over time or with pose, including continuous non-rigid deformation and identifiable parts, joints, or skeletal structures. Non-rigid scenes need not have explicit skeletons."
},
{
"direction_id": "G3",
"name": "Surface recovery and shape completion from 3D samples",
"scope": "Use existing 3D geometric samples, such as points or depth measurements, as primary evidence to recover continuous surfaces or infer substantial unseen shape from incomplete observations."
},
{
"direction_id": "G4",
"name": "View geometry and acquisition-guided mapping",
"scope": "Establish the spatial relationships between views needed for reconstruction and acquire observations as reconstruction proceeds: determine image correspondences, camera poses, and structure; incrementally maintain a map; or select the next observation viewpoint based on the current reconstruction. Offline registration, online mapping, and active acquisition are separate possibilities, not mutual prerequisites."
},
{
"direction_id": "G5",
"name": "Joint recovery of geometry and intrinsic appearance",
"scope": "Jointly recover shape from observations while explicitly decomposing image formation into materials, reflectance, and illumination to obtain faithful or relightable 3D representations."
},
{
"direction_id": "G6",
"name": "Resource-efficient 3D representations",
"scope": "Reduce the practical resource cost of 3D representations while preserving their reconstruction or rendering utility, including training, inference, and rendering computation as well as parameter count and storage. Individual approaches need not optimize both speed and capacity."
},
{
"direction_id": "G7",
"name": "Structured and addressable 3D reconstruction",
"scope": "Recover 3D organization beyond an undivided surface: separable or editable parts and surface properties, connected architectural structures, or scene geometry with semantic and instance labels. Editing, building modeling, and semantic labeling are parallel branches, not jointly required outputs."
},
{
"direction_id": "G8",
"name": "3D recovery from acoustic and radio measurements",
"scope": "Infer 3D shape under sparse or indirect measurement using active-wave returns, such as sonar or radar, as the primary geometric evidence."
},
{
"direction_id": "G9",
"name": "Prompt-conditioned 3D asset creation",
"scope": "Create object or human assets with coherent geometry and appearance from text or other generative prompts, rather than primarily recovering a particular captured subject."
},
{
"direction_id": "G10",
"name": "Reconstruction references and evaluation",
"scope": "Develop resources with aligned observations, reference geometry, or comparative protocols to support evaluation of 3D reconstruction results, rather than contributing a new geometry-estimation algorithm as the primary result."
}
]
}
|