CenterNet : A Machine Learning Model for Anchorless Object Detection

This is an introduction to「CenterNet」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.

Overview

CenterNet is a machine learning model for anchorless object detection published in April 2019.

Objects as Points

CenterNet can be used to calculate the bounding boxes for 80 categories of the COCO dataset.

By using heatmaps, as in other systems such as OpenPose, for object detection, CenterNet can perform detection without using anchors used in YOLOv2 and later.

About anchors

An anchor is a bounding box, defined by several boxes with different aspect ratios. Object detection for each bounding box increases the number of objects that can be detected simultaneously. Introduced in YOLOv2, it increases the number of objects that can be detected simultaneously by performing object detection for each bounding box.

（Source：https://arxiv.org/abs/1904.07850）

Architecture

CenterNet infers a heatmap of the object’s center coordinates, the offset of the center coordinates, and the object’s size.

（Source：https://arxiv.org/abs/1904.07850）

CenterNet performance

CenterNet is capable of more accurate inference than YOLOv3 and RetinaNet.

（Source：https://arxiv.org/abs/1904.07850）

Usage

You can run CenterNet on the webcam video stream in ailia SDK with the following command.

python3 centernet.py -v 0

axinc-ai/ailia-models

ax Inc. has developed ailia SDK, which enables cross-platform, GPU-based rapid inference.

ax Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to contact us for any inquiry.