Downsampled Disentanglement Datasets - Falcor3D and Isaac3D

<p><strong>New disentanglement datasets</strong></p> <p>This data repository contains the&nbsp;<em>Falcor3D</em>&nbsp;and&nbsp;<em>Isaac3D</em>&nbsp;datasets&nbsp;for disentanglement learning,&nbsp;where the image resolution is 128x128.</p> <p><strong><em>Falcor3D</em></strong></p> <p>The Falcor3D dataset consists of&nbsp;<code>233,280</code>&nbsp;images based on the 3D scene of a living room. The meta code corresponds to all possible combinations of&nbsp;<code>7</code>&nbsp;factors of variation:</p> <ul> <li>lighting_intensity (5)</li> <li>lighting_x-dir (6)</li> <li>lighting_y-dir (6)</li> <li>lighting_z-dir (6)</li> <li>camera_x-pos (6)</li> <li>camera_y-pos (6)</li> <li>camera_z-pos (6)</li> </ul> <p>Note that the number&nbsp;<code>m</code>&nbsp;behind each factor represents that the factor has&nbsp;<code>m</code>&nbsp;possible values, uniformly sampled in the normalized range of variations [0, 1].</p> <p>Each image has as filename&nbsp;<em><code>padded_index.png</code></em>&nbsp;where</p> <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> <p>padded_index = index padded with zeros such that it has 6 digits.</p> <p><em><strong>Isaac3D</strong></em></p> <p>The Isaac3D dataset consists of&nbsp;<code>737,280</code>&nbsp;images, based on the 3D scene of a kitchen. The meta code corresponds to all possible combinations of&nbsp;<code>9</code>&nbsp;factors of variation:</p> <ul> <li>object_shape (3)</li> <li>object_scale (4)</li> <li>camera_height (4)</li> <li>robot_x-movement (8)</li> <li>robot_y-movement (5)</li> <li>lighting_intensity (4)</li> <li>lighting_y-dir (6)</li> <li>object_color (4)</li> <li>wall_color (4)</li> </ul> <p>Similarly, the number&nbsp;<code>m</code>&nbsp;behind each factor represents that the factor has&nbsp;<code>m</code>&nbsp;possible values, uniformly sampled in the normalized range of variations [0, 1].</p> <p>Each image has as filename&nbsp;<em><code>padded_index.png</code></em>&nbsp;where</p> <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> <p>padded_index = index padded with zeros such that it has 6 digits.</p> <p>&nbsp;</p>