Cascaded object detection

摘要

Cascaded object detection techniques are described. In one or more implementations, cascaded coarse-to-dense object detection techniques are utilized to detect objects in images. In a first stage, coarse features are extracted from an image, and non-object regions are rejected. Then, in one or more subsequent stages, dense features are extracted from the remaining non-rejected regions of the image to detect one or more objects in the image.

主权项

1. A computer-implemented method comprising:
receiving an image; for a first stage:
extracting, by a processor, coarse features from the image;identifying, by the processor, non-object regions of the image which do not include an object by: computing a confidence score for each region based on the extracted coarse features, comparing the confidence score for each region to a first threshold, and identifying regions with confidence scores that are less than the first threshold as the non-object regions of the image;rejecting the non-object regions of the image; for one or more subsequent stages:
extracting, by the processor, dense features from non-rejected regions of the image;identifying, by the processor, additional non-object regions of the image by computing an additional confidence score for each additional region based on the extracted dense features, comparing the additional confidence score for each additional region to a second threshold, and identifying additional regions with confidence scores that are less than the second threshold as the additional non-object regions of the image, wherein the first threshold is lower than the second threshold;rejecting the additional non-object regions of the image; for a final stage of the one or more subsequent stages, detecting one or more objects in the image based on features extracted in the final stage.