More Like this

1. Multimodality Helps Unimodality: Cross-Modal Few-Shot Learning with Multimodal Models

   Zhiqiu Lin, Samuel Yu (June 2023, CVPR)

   The ability to quickly learn a new task with minimal instruction - known as few-shot learning - is a central aspect of intelligent agents. Classical few-shot benchmarks make use of few-shot samples from a single modality, but such samples may not be sufficient to characterize an entire concept class. In contrast, humans use cross-modal information to learn new concepts efficiently. In this work, we demonstrate that one can indeed build a better visual dog classifier by reading about dogs and listening to them bark. To do so, we exploit the fact that recent multimodal foundation models such as CLIP are inherently cross-modal, mapping different modalities to the same representation space. Specifically, we propose a simple cross-modal adaptation approach that learns from few-shot examples spanning different modalities. By repurposing class names as additional one-shot training samples, we achieve SOTA results with an embarrassingly simple linear classifier for vision-language adaptation. Furthermore, we show that our approach can benefit existing methods such as prefix tuning, adapters, and classifier ensembling. Finally, to explore other modalities beyond vision and language, we construct the first (to our knowledge) audiovisual few-shot benchmark and use cross-modal training to improve the performance of both image and audio classification. 

   more » « less
2. Learning Cross-Modal Audiovisual Representations with Ladder Networks for Emotion Recognition

   https://doi.org/10.1109/ICASSP49357.2023.10096138  

   Goncalves, Lucas; Busso, Carlos (June 2023, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2023))

   Representation learning is a challenging, but essential task in audiovisual learning. A key challenge is to generate strong cross-modal representations while still capturing discriminative information contained in unimodal features. Properly capturing this information is important to increase accuracy and robustness in audio-visual tasks. Focusing on emotion recognition, this study proposes novel cross-modal ladder networks to capture modality-specific in-formation while building strong cross-modal representations. Our method utilizes representations from a backbone network to implement unsupervised auxiliary tasks to reconstruct intermediate layer representations across the acoustic and visual networks. The skip connections between the cross-modal encoder and decoder provide powerful modality-specific and multimodal representations for emotion recognition. Our model on the CREMA-D corpus achieves high performance with precision, recall, and F1 scores over 80% on a six-class problem. 

   more » « less
3. Generalized Visual Odometry via Cross-Modal Self-Training

   Lai, Lei; Shangguan, Zhongkai; Zhang, Jimuyang; Ohn-Bar, Eshed (October 2023, International Conference on Computer Vision)

   We propose XVO, a semi-supervised learning method for training generalized monocular Visual Odometry (VO) models with robust off-the-self operation across diverse datasets and settings. In contrast to standard monocular VO approaches which often study a known calibration within a single dataset, XVO efficiently learns to recover relative pose with real-world scale from visual scene semantics, i.e., without relying on any known camera parameters. We optimize the motion estimation model via self-training from large amounts of unconstrained and heterogeneous dash camera videos available on YouTube. Our key contribution is twofold. First, we empirically demonstrate the benefits of semi-supervised training for learning a general-purpose direct VO regression network. Second, we demonstrate multi-modal supervision, including segmentation, flow, depth, and audio auxiliary prediction tasks, to facilitate generalized representations for the VO task. Specifically, we find audio prediction task to significantly enhance the semi-supervised learning process while alleviating noisy pseudo-labels, particularly in highly dynamic and out-of-domain video data. Our proposed teacher network achieves state-of-the-art performance on the commonly used KITTI benchmark despite no multi-frame optimization or knowledge of camera parameters. Combined with the proposed semi-supervised step, XVO demonstrates off-the-shelf knowledge transfer across diverse conditions on KITTI, nuScenes, and Argoverse without fine-tuning. 

   more » « less
4. Multimodaltrace: Deepfake Detection using Audiovisual Representation Learning

   Muhammad Anas Raza Khalid Mahmood Malik (June 2023, IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Vancouver, CA, 2023)

   By employing generative deep learning techniques, Deepfakes are created with the intent to create mistrust in society, manipulate public opinion and political decisions, and for other malicious purposes such as blackmail, scamming, and even cyberstalking. As realistic deepfake may involve manipulation of either audio or video or both, thus it is important to explore the possibility of detecting deepfakes through the inadequacy of generative algorithms to synchronize audio and visual modalities. Prevailing performant methods, either detect audio or video cues for deepfakes detection while few ensemble the results after predictions on both modalities without inspecting relationship between audio and video cues. Deepfake detection using joint audiovisual representation learning is not explored much. Therefore, this paper proposes a unified multimodal framework, Multimodaltrace, which extracts learned channels from audio and visual modalities, mixes them independently in IntrAmodality Mixer Layer (IAML), processes them jointly in IntErModality Mixer Layers (IEML) from where it is fed to multilabel classification head. Empirical results show the effectiveness of the proposed framework giving state-of-the-art accuracy of 92.9% on the FakeAVCeleb dataset. The cross-dataset evaluation of the proposed framework on World Leaders and Presidential Deepfake Detection Datasets gives an accuracy of 83.61% and 70% respectively. The study also provides insights into how the model focuses on different parts of audio and visual features through integrated gradient analysis 

   more » « less
5. Action2Sound: Ambient-Aware Generation of Action Sounds from Egocentric Videos

   Chen, Changan; Peng, Puyuan; Baid, Ami; Xue, Zihui; Hsu, Wei-Ning; Harwath, David; Grauman, Kristen (July 2024, https://doi.org/10.48550/arXiv.2406.09272)

   Generating realistic audio for human actions is critical for applications such as film sound effects and virtual reality games. Existing methods assume complete correspondence between video and audio during training, but in real-world settings, many sounds occur off-screen or weakly correspond to visuals, leading to uncontrolled ambient sounds or hallucinations at test time. This paper introduces AV-LDM, a novel ambient-aware audio generation model that disentangles foreground action sounds from ambient background noise in in-the-wild training videos. The approach leverages a retrieval-augmented generation framework to synthesize audio that aligns both semantically and temporally with the visual input. Trained and evaluated on Ego4D and EPIC-KITCHENS datasets, along with the newly introduced Ego4D-Sounds dataset (1.2M curated clips with action-audio correspondence), the model outperforms prior methods, enables controllable ambient sound generation, and shows promise for generalization to synthetic video game clips. This work is the first to emphasize faithful video-to-audio generation focused on observed visual content despite noisy, uncurated training data. 

   more » « less