From images in the wild to video-informed image classification
This addresses the challenge of improving image classification accuracy for complex real-world images, particularly in domain-specific applications like plant identification, though it appears incremental as it builds on existing classifiers with video enhancements.
The paper tackled the problem of image classifiers failing on images with high visual complexity by applying state-of-the-art object classifiers to a unique dataset from Bali and introducing a novel approach that combines video cues with an ensemble of imperfect classifiers to improve classification results on video-sourced plant images.
Image classifiers work effectively when applied on structured images, yet they often fail when applied on images with very high visual complexity. This paper describes experiments applying state-of-the-art object classifiers toward a unique set of images in the wild with high visual complexity collected on the island of Bali. The text describes differences between actual images in the wild and images from Imagenet, and then discusses a novel approach combining informational cues particular to video with an ensemble of imperfect classifiers in order to improve classification results on video sourced images of plants in the wild.