More Like this

1. VideoSSL: Semi-Supervised Learning for Video Classification

   Jing, L; Parag, T; Wu, Z; Tian, Y; Wang, H. (January 2021, Winter Conference on Applications of Computer Vision (WACV), 2021.)

   null (Ed.)

   We propose a semi-supervised learning approach for video classification, VideoSSL, using convolutional neural networks (CNN). Like other computer vision tasks, existing supervised video classification methods demand a large amount of labeled data to attain good performance. However, annotation of a large dataset is expensive and time consuming. To minimize the dependence on a large annotated dataset, our proposed semi-supervised method trains from a small number of labeled examples and exploits two regulatory signals from unlabeled data. The first signal is the pseudo-labels of unlabeled examples computed from the confidences of the CNN being trained. The other is the normalized probabilities, as predicted by an image classifier CNN, that captures the information about appearances of the interesting objects in the video. We show that, under the supervision of these guiding signals from unlabeled examples, a video classification CNN can achieve impressive performances utilizing a small fraction of annotated examples on three publicly available datasets: UCF101, HMDB51, and Kinetics. 

   more » « less
2. Self-Supervised Learning for Online Anomaly Detection in High-Dimensional Data Streams

   https://doi.org/10.3390/electronics12091971  

   Mozaffari, Mahsa; Doshi, Keval; Yilmaz, Yasin (May 2023, Electronics)

   In this paper, we address the problem of detecting and learning anomalies in high-dimensional data-streams in real-time. Following a data-driven approach, we propose an online and multivariate anomaly detection method that is suitable for the timely and accurate detection of anomalies. We propose our method for both semi-supervised and supervised settings. By combining the semi-supervised and supervised algorithms, we present a self-supervised online learning algorithm in which the semi-supervised algorithm trains the supervised algorithm to improve its detection performance over time. The methods are comprehensively analyzed in terms of computational complexity, asymptotic optimality, and false alarm rate. The performances of the proposed algorithms are also evaluated using real-world cybersecurity datasets, that show a significant improvement over the state-of-the-art results. 

   more » « less
3. Multiview Supervision By Registration

   Zhang, Y.; Park, H.S. (January 2020, IEEE Winter Conference on Applications of Computer Vision)

   This paper presents a semi-supervised learning framework to train a keypoint detector using multiview image streams given the limited number of labeled instances (typically <4%). We leverage three self-supervisionary signals in multiview tracking to utilize the unlabeled data: (1) a keypoint in one view can be supervised by other views via epipolar geometry; (2) a keypoint detection must be consistent across time; (3) a visible keypoint in one view is likely to be visible in the adjacent view. We design a new end-toend network that can propagate these self-supervisionary signals across the unlabeled data from the labeled data in a differentiable manner. We show that our approach outperforms existing detectors including DeepLabCut tailored to the keypoint detection of non-human species such as monkeys, dogs, and mice. 

   more » « less
4. w-HAR: An Activity Recognition Dataset and Framework Using Low-Power Wearable Devices

   https://doi.org/10.3390/s20185356  

   Bhat, Ganapati; Tran, Nicholas; Shill, Holly; Ogras, Umit Y. (September 2020, Sensors)

   null (Ed.)

   Human activity recognition (HAR) is growing in popularity due to its wide-ranging applications in patient rehabilitation and movement disorders. HAR approaches typically start with collecting sensor data for the activities under consideration and then develop algorithms using the dataset. As such, the success of algorithms for HAR depends on the availability and quality of datasets. Most of the existing work on HAR uses data from inertial sensors on wearable devices or smartphones to design HAR algorithms. However, inertial sensors exhibit high noise that makes it difficult to segment the data and classify the activities. Furthermore, existing approaches typically do not make their data available publicly, which makes it difficult or impossible to obtain comparisons of HAR approaches. To address these issues, we present wearable HAR (w-HAR) which contains labeled data of seven activities from 22 users. Our dataset’s unique aspect is the integration of data from inertial and wearable stretch sensors, thus providing two modalities of activity information. The wearable stretch sensor data allows us to create variable-length segment data and ensure that each segment contains a single activity. We also provide a HAR framework to use w-HAR to classify the activities. To this end, we first perform a design space exploration to choose a neural network architecture for activity classification. Then, we use two online learning algorithms to adapt the classifier to users whose data are not included at design time. Experiments on the w-HAR dataset show that our framework achieves 95% accuracy while the online learning algorithms improve the accuracy by as much as 40%. 

   more » « less
5. Extending the WILDS Benchmark for Unsupervised Adaptation

   Sagawa, Shiori; Koh, Pang Wei; Lee, Tony; Gao, Irene; Xie, Sang Michael; Shen, Kendrick; Kumar, Ananya; Hu, Weihua; Yasunaga, Michihiro; Marklund, Henrik; et al (January 2022, International Conference on Learning Representations (ICLR))

   Machine learning systems deployed in the wild are often trained on a source distribution but deployed on a different target distribution. Unlabeled data can be a powerful point of leverage for mitigating these distribution shifts, as it is frequently much more available than labeled data and can often be obtained from distributions beyond the source distribution as well. However, existing distribution shift benchmarks with unlabeled data do not reflect the breadth of scenarios that arise in real-world applications. In this work, we present the WILDS 2.0 update, which extends 8 of the 10 datasets in the WILDS benchmark of distribution shifts to include curated unlabeled data that would be realistically obtainable in deployment. These datasets span a wide range of applications (from histology to wildlife conservation), tasks (classification, regression, and detection), and modalities (photos, satellite images, microscope slides, text, molecular graphs). The update maintains consistency with the original WILDS benchmark by using identical labeled training, validation, and test sets, as well as identical evaluation metrics. We systematically benchmark state-of-the-art methods that use unlabeled data, including domain-invariant, self-training, and self-supervised methods, and show that their success on WILDS is limited. To facilitate method development, we provide an open-source package that automates data loading and contains the model architectures and methods used in this paper. 

   more » « less