Learning with privileged information via adversarial discriminative modality distillation
This addresses the challenge of deploying models in real-life scenarios with missing modalities, specifically for RGB-D vision tasks, though it is incremental as it builds on existing adversarial and privileged information frameworks.
The paper tackles the problem of learning from multimodal training data (RGB-D) when only a single modality (RGB) is available at test time, by proposing an adversarial discriminative modality distillation method that achieves state-of-the-art results on object classification and video action recognition datasets.
Heterogeneous data modalities can provide complementary cues for several tasks, usually leading to more robust algorithms and better performance. However, while training data can be accurately collected to include a variety of sensory modalities, it is often the case that not all of them are available in real life (testing) scenarios, where a model has to be deployed. This raises the challenge of how to extract information from multimodal data in the training stage, in a form that can be exploited at test time, considering limitations such as noisy or missing modalities. This paper presents a new approach in this direction for RGB-D vision tasks, developed within the adversarial learning and privileged information frameworks. We consider the practical case of learning representations from depth and RGB videos, while relying only on RGB data at test time. We propose a new approach to train a hallucination network that learns to distill depth information via adversarial learning, resulting in a clean approach without several losses to balance or hyperparameters. We report state-of-the-art results on object classification on the NYUD dataset and video action recognition on the largest multimodal dataset available for this task, the NTU RGB+D, as well as on the Northwestern-UCLA.