A High-Performance Object Proposals based on Horizontal High Frequency Signal
This work addresses a key bottleneck in object detection preprocessing for computer vision applications, offering a significant performance improvement.
The authors tackled the challenge of balancing detection recall, localization quality, and computational efficiency in object proposal methods by proposing BIHL, which achieved the highest detection recall and 79.5% mean average best overlap on VOC2007 while being nearly three times faster than the fastest existing method.
In recent years, the use of object proposal as a preprocessing step for target detection to improve computational efficiency has become an effective method. Good object proposal methods should have high object detection recall rate and low computational cost, as well as good localization quality and repeatability. However, it is difficult for current advanced algorithms to achieve a good balance in the above performance. For this problem, we propose a class-independent object proposal algorithm BIHL. It combines the advantages of window scoring and superpixel merging, which not only improves the localization quality but also speeds up the computational efficiency. The experimental results on the VOC2007 data set show that when the IOU is 0.5 and 10,000 budget proposals, our method can achieve the highest detection recall and an mean average best overlap of 79.5%, and the computational efficiency is nearly three times faster than the current fastest method. Moreover, our method is the method with the highest average repeatability among the methods that achieve good repeatability to various disturbances.