Figure 2From: Automated detection of elephants in wildlife video Overview of automated elephant detection. First, a color model is generated from labeled ground truth images. Next, image segments are classified by the color model. Positively detected segments (candidates) are tracked through the sequence resulting in spatiotemporally coherent candidates. The final detections are obtained by validating the spatiotemporal candidates by shape, texture, and consistency constraints. Finally, postprocessing fills gaps in tracking for each detection.Back to article page