Fig. 13. Photorealistic reconstruction for unconstrained indoor scenes.
(a) The PhlatCam-Webcam setup to capture the dataset for finetuning FlatNet-gen. (b) Tikhonov reconstruction. (c) Reconstructions from FlatNet-gen trained just on display captured data. (d) Reconstructions using FlatNet-gen finetuned on unconstrained indoor captures. (e) Webcam image for reference. Finetuning makes the reconstructions more realistic.