Downsampled Disentanglement Datasets - Falcor3D and Isaac3D

<strong>New disentanglement datasets</strong> This data repository contains the <em>Falcor3D</em> and <em>Isaac3D</em> datasets for disentanglement learning, where the image resolution is 128x128. <strong><em>Falcor3D</em></strong> The Falcor3D dataset consists of <code>233,280</code> images based on the 3D scene of a living room. The meta code corresponds to all possible combinations of <code>7</code> factors of variation: lighting_intensity (5) lighting_x-dir (6) lighting_y-dir (6) lighting_z-dir (6) camera_x-pos (6) camera_y-pos (6) camera_z-pos (6) Note that the number <code>m</code> behind each factor represents that the factor has <code>m</code> possible values, uniformly sampled in the normalized range of variations [0, 1]. Each image has as filename <em><code>padded_index.png</code></em> where <pre><code>index = lighting_intensity * 46656 + lighting_x-dir * 7776 + lighting_y-dir * 1296 + lighting_z-dir * 216 + camera_x-pos * 36 + camera_y-pos * 6 + camera_z-pos </code></pre> padded_index = index padded with zeros such that it has 6 digits. <em><strong>Isaac3D</strong></em> The Isaac3D dataset consists of <code>737,280</code> images, based on the 3D scene of a kitchen. The meta code corresponds to all possible combinations of <code>9</code> factors of variation: object_shape (3) object_scale (4) camera_height (4) robot_x-movement (8) robot_y-movement (5) lighting_intensity (4) lighting_y-dir (6) object_color (4) wall_color (4) Similarly, the number <code>m</code> behind each factor represents that the factor has <code>m</code> possible values, uniformly sampled in the normalized range of variations [0, 1]. Each image has as filename <em><code>padded_index.png</code></em> where <pre><code>index = object_shape * 245760 + object_scale * 30720 + camera_height * 6144 + robot_x-movement * 1536 + robot_y-movement * 384 + lighting_intensity * 96 + lighting_y-dir * 16 + object_color * 4 + wall color </code></pre> padded_index = index padded with zeros such that it has 6 digits.<strong>新解耦数据集</strong> 本数据集仓库包含用于解耦学习（disentanglement learning）的<em>Falcor3D</em>与<em>Isaac3D</em>数据集，图像分辨率均为128×128。 <strong><em>Falcor3D</em></strong> Falcor3D数据集基于客厅3D场景，共包含233280张图像。其元编码对应7个变化因子的所有可能组合：光照强度（lighting_intensity，5种取值）、光照x方向（lighting_x-dir，6种取值）、光照y方向（lighting_y-dir，6种取值）、光照z方向（lighting_z-dir，6种取值）、相机x位置（camera_x-pos，6种取值）、相机y位置（camera_y-pos，6种取值）、相机z位置（camera_z-pos，6种取值）。需注意，每个因子后的数字m代表该因子在归一化变化范围[0, 1]内均匀采样得到m种可能取值。每张图像的文件名为<em><code>padded_index.png</code></em>，其中： <pre><code>index = lighting_intensity * 46656 + lighting_x-dir * 7776 + lighting_y-dir * 1296 + lighting_z-dir * 216 + camera_x-pos * 36 + camera_y-pos * 6 + camera_z-pos</code></pre> `padded_index`为将index补零至6位数字后的结果。 <strong><em>Isaac3D</em></strong> Isaac3D数据集基于厨房3D场景，共包含737280张图像。其元编码对应9个变化因子的所有可能组合：物体形状（object_shape，3种取值）、物体缩放比例（object_scale，4种取值）、相机高度（camera_height，4种取值）、机器人x方向移动量（robot_x-movement，8种取值）、机器人y方向移动量（robot_y-movement，5种取值）、光照强度（lighting_intensity，4种取值）、光照y方向（lighting_y-dir，6种取值）、物体颜色（object_color，4种取值）、墙面颜色（wall_color，4种取值）。同理，每个因子后的数字m代表该因子在归一化变化范围[0, 1]内均匀采样得到m种可能取值。每张图像的文件名为<em><code>padded_index.png</code></em>，其中： <pre><code>index = object_shape * 245760 + object_scale * 30720 + camera_height * 6144 + robot_x-movement * 1536 + robot_y-movement * 384 + lighting_intensity * 96 + lighting_y-dir * 16 + object_color * 4 + wall_color</code></pre> `padded_index`为将index补零至6位数字后的结果。