Algorithm 1: Synthetic dataset generator.
Input: CAD models of empty furniture $M_{CAD}$ and $clutterset$ Output: furniture model depth image $I_{3D}$ with set of annotated oriented bounding boxes of storage volumes $OBB$
procedure Voxelization($M_{CAD}$)	▹ Process I
Transform $M_{CAD}$ from mesh to voxel sapce     return $M_{voxel}$ end procedure
procedure Empty storage volume detection($M_{voxel}$)	▹ Process I
Step 1. Compute occupancy grid     Step 2. Compute occupancy distances     Step 3. Detect free space cores     Step 4. Detect empty storage volumes $ESV$     Step 5. Prune empty storage volumes     Step 6. Transform empty storage volumes from voxel to mesh space     Step 7. Adjust detected axis-aligned bounding boundig boxes $AABB$ to enclosed storage volumes     return $AABB$ end procedure
procedure Synthetic scene generator($M_{CAD},AABB,clutterset$)	▹ Process II
Step 1. Scale $M_{CAD}$ using a random scale     Step 2. Add random items from $clutterset$ to each $AABB$     Step 3. Place $M_{CAD}$ on a horizontal plane representing the floor     Step 4. With $P\left(wall\right)$ generate vertical plane behind $M_{CAD}$ representing a wall     Step 5. Position virtual camera at $T_{PoV}$ facing $M_{CAD}$     Step 6. Render depth images $I_{3D}$     Step 7. Superpose simulated camera noise to $I_{3D}$     Step 8. Get oriented bounding boxes $OBB$ from $AABB$ and $C_{PoV}$     return $I_{3D},OBB$ end procedure