Determine which categories does each object belong to

Filters every node in the House based on a set of node_filters Then filters every room (with nodes filtered already) in the House bas

Represents a House See SUNCG Toolbox for a detailed description of the structure of the json file describing a house

Standard vertex-face representation, triangulated Order: x, z, y

Various information associated with the objects

Representing a node in synthesis time

Loading a rendered room Attributes ---------- category_map (ObjectCategories): object category mapping that should be the sa

Provides observation information for a collection of objects

Represents an affine transformation composed of translation, rotation, scale

Generates projection between original 3D space and rendered 2D image space Given the position of the room

Dataset for training the baseline NADE model for predicting object category counts

Converts the dataset back to the original json format

Dataset for training/testing the location/category network

Encapsulates a set of object arrangement priors

Dataset for training/testing the instance/orientation network

Dataset for training/testing the "should continue" network

Iterative optimization of object arrangements using greedy sampling of ArrangementPriors

A collection of object arrangement priors for estimating probability of an arrangement and sampling arrangements

Class that processed the pickled version of the SUNCG dataset

Pre-render top-down view of each room in the house (floor, walls and objects)

Save the (usually filtered) dataset into pickle files

Gather stats of the houses Useful to get an idea of what's in the dataset And must be called to generate the model frequency information

Represents a floor level in the house Currently mostly just used to parse the list of nodes and rooms Might change in the future

Basic unit of representation of SUNCG Usually a room or an object Refer to SUNCG toolbox for the possible set of attributes

Encapsulates arrangement priors for a specific category key (room_type, obj_category, ref_obj_category)

Multi-channel top-down composite, used as input to NN

Represents a room in the house

Class that synthesizes scenes based on the trained models

Class that synthesize a single room and keeps its record

Take a room, pre-render top-down views Of floor, walls and individual objects That can be used to generate the multi-channel views used i

Abstract base class defining actions that can be done to the Dataset