# Automatic detection of boosted Higgs boson and top quark jets in an event image

## TL;DR

This paper presents a deep neural network based on Mask R-CNN for detecting Higgs and top quark jets in event images, improving tagging and momentum prediction with robustness across different event types.

## Contribution

The authors introduce a novel supervised training algorithm, a new jet branch for constituent-based momentum prediction, and demonstrate superior performance over existing models.

## Key findings

- Outperforms LorentzNet in jet tagging accuracy.
- Surpasses PELICAN in momentum regression.
- Maintains performance across different event processes and overlapping jets.

## Abstract

We build a deep neural network based on the Mask R-CNN framework to detect the Higgs jets and top quark jets in any event image. We propose an algorithm to assign the top quark final states at the ground truth level so that the network can be trained in a supervised manner. A new jet branch is added to the network, which uses constituent information to predict the four-momenta of the original parton, thus intrinsically implementing the pileup mitigation. The network can predict both the shapes and the momenta of target jets. We show that the network surpasses the LorentzNet in top and Higgs tagging and the PELICAN network in momentum regression for certain cases, in terms of reconstruction efficiency and accuracy. We also show that the performance of the network does not degrade much when applied to events of a process different from the trained one and to events with overlapping jets.