Intrinsic Image Decomposition using Paradigms

TL;DR

This paper introduces an unsupervised approach for intrinsic image decomposition that does not require annotated data, yet achieves competitive WHDR scores by leveraging paradigms and a novel smoothing technique.

Contribution

It presents a novel unsupervised method for intrinsic image decomposition that avoids reliance on ground truth or rendered data, using paradigms and smoothing for effective results.

Findings

Achieves competitive WHDR scores without supervised data.

Provides estimates of test/train variance in WHDR scores.

Demonstrates the feasibility of unsupervised intrinsic image decomposition.

Abstract

Intrinsic image decomposition is the classical task of mapping image to albedo. The WHDR dataset allows methods to be evaluated by comparing predictions to human judgements ("lighter", "same as", "darker"). The best modern intrinsic image methods learn a map from image to albedo using rendered models and human judgements. This is convenient for practical methods, but cannot explain how a visual agent without geometric, surface and illumination models and a renderer could learn to recover intrinsic images. This paper describes a method that learns intrinsic image decomposition without seeing WHDR annotations, rendered data, or ground truth data. The method relies on paradigms - fake albedos and fake shading fields - together with a novel smoothing procedure that ensures good behavior at short scales on real images. Long scale error is controlled by averaging. Our method achieves WHDR scores competitive with those of strong recent methods allowed to see training WHDR annotations, rendered data, and ground truth data. Because our method is unsupervised, we can compute estimates of the test/train variance of WHDR scores; these are quite large, and it is unsafe to rely small differences in reported WHDR.