TalBarami/msd_3dshapes
MSD Shapes3D Dataset Attribution The Multi-factor Sequential Disentanglement benchmark includes a modified variant of the 3D Shapes dataset, adapted to support sequential multi-factor disentanglement. In this setup, the floor hue, wall hue, object hue, and shape are static, while the scale and orientation change over time. Original repository:https://github.com/deepmind/3d-shapes @misc{3dshapes18, title={3D Shapes Dataset}, author={Burgess, Chris and Kim, Hyunjik}… See the full description on the dataset page: https://huggingface.co/datasets/TalBarami/msd_3dshapes.
021
1{2 "floor_hue": {3 "index": 0,4 "type": "static",5 "n_classes": 10,6 "ignore": false,7 "values": {8 "0.0": 0,9 "0.1": 1,10 "0.2": 2,11 "0.3": 3,12 "0.4": 4,13 "0.5": 5,14 "0.6": 6,15 "0.7": 7,16 "0.8": 8,17 "0.9": 918 }19 },20 "wall_hue": {21 "index": 1,22 "type": "static",23 "n_classes": 10,24 "ignore": false,25 "values": {26 "0.0": 0,27 "0.1": 1,28 "0.2": 2,29 "0.3": 3,30 "0.4": 4,31 "0.5": 5,32 "0.6": 6,33 "0.7": 7,34 "0.8": 8,35 "0.9": 936 }37 },38 "object_hue": {39 "index": 2,40 "type": "static",41 "n_classes": 10,42 "ignore": false,43 "values": {44 "0.0": 0,45 "0.1": 1,46 "0.2": 2,47 "0.3": 3,48 "0.4": 4,49 "0.5": 5,50 "0.6": 6,51 "0.7": 7,52 "0.8": 8,53 "0.9": 954 }55 },56 "shape": {57 "index": 3,58 "type": "static",59 "n_classes": 4,60 "ignore": false,61 "values": {62 "cube": 0,63 "cylinder": 1,64 "sphere": 2,65 "capsule": 366 }67 },68 "scale_dynamic": {69 "index": 4,70 "type": "dynamic",71 "n_classes": 6,72 "ignore": false,73 "values": {74 "increasing_1x": 0,75 "decreasing_1x": 1,76 "increasing_2x": 2,77 "decreasing_2x": 3,78 "alternating_big_small": 4,79 "alternating_small_big": 580 }81 },82 "orientation_dynamic": {83 "index": 5,84 "type": "dynamic",85 "n_classes": 3,86 "ignore": false,87 "values": {88 "counterclockwise": 0,89 "static": 1,90 "clockwise": 291 }92 }93}