DiverseNet: When One Right Answer is not Enough

TL;DR

DiverseNet introduces a simple neural network training method that enables multiple plausible structured predictions per input, improving diversity and accuracy in tasks like image completion and flow prediction.

Contribution

It presents a novel, easy-to-implement approach for training models to produce diverse outputs, addressing mode collapse and outperforming ensemble-based methods.

Findings

Improves diversity in structured predictions across multiple tasks

Outperforms ensemble methods in accuracy and speed

Achieves state-of-the-art results in image completion, volume estimation, and flow prediction

Abstract

Many structured prediction tasks in machine vision have a collection of acceptable answers, instead of one definitive ground truth answer. Segmentation of images, for example, is subject to human labeling bias. Similarly, there are multiple possible pixel values that could plausibly complete occluded image regions. State-of-the art supervised learning methods are typically optimized to make a single test-time prediction for each query, failing to find other modes in the output space. Existing methods that allow for sampling often sacrifice speed or accuracy. We introduce a simple method for training a neural network, which enables diverse structured predictions to be made for each test-time query. For a single input, we learn to predict a range of possible answers. We compare favorably to methods that seek diversity through an ensemble of networks. Such stochastic multiple choice learning faces mode collapse, where one or more ensemble members fail to receive any training signal. Our best performing solution can be deployed for various tasks, and just involves small modifications to the existing single-mode architecture, loss function, and training regime. We demonstrate that our method results in quantitative improvements across three challenging tasks: 2D image completion, 3D volume estimation, and flow prediction.