Publications

Poster: Time-Efficient Sparse and Lightweight Adaptation for Real-Time Mobile Applications
Poster: Time-Efficient Sparse and Lightweight Adaptation for Real-Time Mobile Applications

We propose TESLA Time-Efficient Sparse and Lightweight Adaptation strategy for real-time mobile applications, which skips adaptation for specific batches to increase the inference sample rate. Our method balances model accuracy and inference speed by accumulating domain-informative samples from non-adapted batches and sparsely adapting them.

Jun 3, 2024

IMG2IMU: Translating Knowledge from Large-Scale Images to IMU Sensing Applications
IMG2IMU: Translating Knowledge from Large-Scale Images to IMU Sensing Applications

Pre-training representations acquired via self-supervised learning could achieve high accuracy on even tasks with small training data. Unlike in vision and natural language processing domains, pre-training for IMU-based applications is challenging, as there are few public datasets with sufficient size and diversity to learn generalizable representations. To overcome this problem, we propose IMG2IMU that adapts pre-trained representation from large-scale images to diverse IMU sensing tasks. We convert the sensor data into visually interpretable spectrograms for the model to utilize the knowledge gained from vision. We further present a sensor-aware pre-training method for images that enables models to acquire particularly impactful knowledge for IMU sensing applications. This involves using contrastive learning on our augmentation set customized for the properties of sensor data. Our evaluation with four different IMU sensing tasks shows that IMG2IMU outperforms the baselines pre-trained on sensor data by an average of 9.6%p F1-score, illustrating that vision knowledge can be usefully incorporated into IMU sensing applications where only limited training data is available.

Feb 29, 2024

Sherlock: Automated Hidden Camera Detection with Shutter Speed Adaptation
Sherlock: Automated Hidden Camera Detection with Shutter Speed Adaptation

With the rapid advancements in surveillance camera technology, there has been a surge in crimes involving illegal filming using hidden cameras. However, current solutions necessitate slow manual scanning processes that require thousands ofworkers within a city to ensure coverage. To address this challenge, we introduce Sherlock, a fully automated hidden camera detection system that utilizes a drone in combination with a flickering flashlight. To expedite the scanning process and accommodate the irregular movements of the drone, Sherlock leverages the rolling shutter effect to capture images when the flashlight is turned on and off within a single frame. This methodology enables Sherlock to efficiently identify reflective objects and detect hidden camera lenses among them. To demonstrate the feasibility of Sherlock, we present a proof-of-concept evaluation by deploying a drone in various environments to detect and locate hidden cameras.

Oct 25, 2023