VideoPose: Estimating 6D object pose from videos

Abstract

We introduce a simple yet effective algorithm that uses convolutional neural networks to directly estimate object poses from videos. Our approach leverages the temporal information from a video sequence, and is computationally efficient and robust to support robotic and AR domains. Our proposed network takes a pre-trained 2D object detector as input, and aggregates visual features through a recurrent neural network to make predictions at each frame. Experimental evaluation on the YCB-Video dataset show that our approach is on par with the state-of-the-art algorithms. Further, with a speed of 30 fps, it is also more efficient than the state-of-the-art, and therefore applicable to a variety of applications that require real-time object pose estimation.

Exploring the potential of the sit-to-stand test for self-assessment of physical condition in advanced knee osteoarthritis patients using computer vision

“…Furthermore, The VideoPose approach, an open-source video-based three-dimensional pose estimation method, simplifies the complexities of learning from two-dimensional data. This method enables the acquisition of three-dimensional key body positions without requiring intricate feature learning processes, thanks to its foundation on time-dilated convolutional networks, which exhibit strong performance in dynamic environments ( 30–32 ). The integration of these methods empowers the extraction and comprehensive analysis of spatiotemporal motion insights derived from camera-generated data.…”