It's About Time: Analog Clock Reading in the Wild
This addresses the challenge of automated time reading from analog clocks in real-world scenarios, with incremental improvements in reducing annotation effort through synthetic data and pseudo-labeling.
The paper tackles the problem of reading analog clocks in natural images or videos by developing a framework that uses synthetic data generation and a spatial transformer network for training, achieving good accuracy on real clocks and introducing three benchmark datasets with full time annotations.
In this paper, we present a framework for reading analog clocks in natural images or videos. Specifically, we make the following contributions: First, we create a scalable pipeline for generating synthetic clocks, significantly reducing the requirements for the labour-intensive annotations; Second, we introduce a clock recognition architecture based on spatial transformer networks (STN), which is trained end-to-end for clock alignment and recognition. We show that the model trained on the proposed synthetic dataset generalises towards real clocks with good accuracy, advocating a Sim2Real training regime; Third, to further reduce the gap between simulation and real data, we leverage the special property of "time", i.e.uniformity, to generate reliable pseudo-labels on real unlabelled clock videos, and show that training on these videos offers further improvements while still requiring zero manual annotations. Lastly, we introduce three benchmark datasets based on COCO, Open Images, and The Clock movie, with full annotations for time, accurate to the minute.