MBZUAI researchers presented a new approach to video analysis at ICCV in Paris, led by Syed Talal Wasim. The approach builds on still image processing techniques like focal modulation to analyze spatial and temporal information in video separately. It aims to improve temporal aggregation while avoiding the computational complexity of transformers. Why it matters: This research advances video understanding in computer vision by offering a more efficient method for temporal modeling, crucial for applications like activity recognition and video surveillance.
Researchers at KAUST have developed a new method called Deep State Identifier for extracting information from videos for reinforcement learning. The method learns to predict returns from video-encoded episodes and identifies critical states using mask-based sensitivity analysis. Experiments demonstrate the method's potential for understanding and improving agent behavior in DRL.
MBZUAI researchers presented a new approach to video question answering at ICCV 2023. The method leverages insights from analyzing still images to understand video content, potentially reducing the computational resources needed for training video question answering models. Guangyi Chen, Kun Zhang, and colleagues aim to apply pre-trained image models to understand video concepts. Why it matters: This research could lead to more efficient and accessible video analysis tools, benefiting fields like healthcare and security where video data is abundant.
MBZUAI researchers introduce VideoMathQA, a new benchmark for evaluating mathematical reasoning in videos, requiring models to interpret visual information, text, and spoken cues. The dataset spans 10 mathematical domains with videos ranging from 10 seconds to over 1 hour, and includes multi-step reasoning annotations. The benchmark aims to evaluate temporal cross-modal reasoning and highlights the limitations of existing approaches in complex video-based mathematical problem solving.
KAUST's Image and Video Understanding Lab is developing machine learning algorithms for computer vision and object tracking, with applications in video content search and UAV navigation. Their algorithms can detect specific activities in videos, helping platforms detect unwanted content and deliver relevant ads. The object tracking algorithm is also used to empower UAVs, enabling them to follow objects autonomously. Why it matters: This research enhances video content analysis and UAV capabilities, positioning KAUST as a leader in computer vision and AI applications within the region.