Bottom-up Object Detection by Grouping Extreme and Center Points

TL;DR

This paper presents a bottom-up object detection method using keypoints for extreme and center points, achieving competitive accuracy with state-of-the-art region-based methods and providing better segmentation masks.

Contribution

It introduces a novel bottom-up approach that detects objects via keypoints, eliminating the need for region classification, and demonstrates improved segmentation masks.

Findings

Achieves 43.2% bounding box AP on COCO test-dev.

Provides coarse octagonal masks with 18.9% Mask AP.

Enhances segmentation to 34.6% Mask AP with extreme point guidance.

Abstract

With the advent of deep learning, object detection drifted from a bottom-up to a top-down recognition problem. State of the art algorithms enumerate a near-exhaustive list of object locations and classify each into: object or not. In this paper, we show that bottom-up approaches still perform competitively. We detect four extreme points (top-most, left-most, bottom-most, right-most) and one center point of objects using a standard keypoint estimation network. We group the five keypoints into a bounding box if they are geometrically aligned. Object detection is then a purely appearance-based keypoint estimation problem, without region classification or implicit feature learning. The proposed method performs on-par with the state-of-the-art region based detection methods, with a bounding box AP of 43.2% on COCO test-dev. In addition, our estimated extreme points directly span a coarse octagonal mask, with a COCO Mask AP of 18.9%, much better than the Mask AP of vanilla bounding boxes. Extreme point guided segmentation further improves this to 34.6% Mask AP.