Cascaded refinement network (CRN) block, as described in: Qifeng Chen and Vladlen Koltun, "Photographic Image Synthesis with Cascaded Refinemen

ResNet block used in SPADE. It differs from the ResNet block of pix2pixHD in that it takes in the segmentation map as input, learns the s

Code to load all the frames and associated metadata in a video

DS Head to predict downstream task from graph Args: num_classes (int) gnn_cfg (ConfigType): gnn cfg img_feat_key (str): u

Decoder Network that generates a target image from a pair of masked source image and layout Implemented in two options: with a CRN block or a

Detector that also outputs downstream predictions. Args: detector (ConfigType): underlying object detector config num_classes: in

Temporal extension of DeepCVS. replace decoder predictor with transformer.

GNN module that takes a graph (BaseDataElement) object and processes the node and edge features. Args: num_layers (int) arch

Graph Head to construct graph from detections Args: edges_per_node (int) viz_feat_size (int) roi_extractor gnn_cf

Detector that also outputs a scene graph, reconstructed image, and any downstream predictions. Args: detector (ConfigType): underlying ob

Reconstruction Head from object-centric features Args: img_decoder_cfg (ConfigType): decoder img_aspect_ratio (tuple[int]): aspe

Extract RoI features from a single level feature map. If there are multiple input feature levels, each RoI is mapped to a level according to

Code to sample a keyframe from the entire dataset

Code to load all images and metadata for a clip given a keyframe