Andreas Petridis | Searcharxiv

arXiv2023

Fashion Object Detection for Tops & Bottoms

Fashion is one of the largest world's industries and computer vision techniques have been becoming more popular in recent years, in particular, for tasks such as object detection and apparel segmentation. Even with the rapid growth in computer vision solutions, specifically for the fashion industry, many problems are far for being resolved. Therefore, not at all times, adjusting out-of-the-box pre-trained computer vision models will provide the desired solution. In the present paper is proposed a pipeline that takes a noisy image with a person and specifically detects the regions with garments that are bottoms or tops. Our solution implements models that are capable of finding human parts in an image e.g. full-body vs half-body, or no human is found. Then, other models knowing that there's a human and its composition (e.g. not always we have a full-body) finds the bounding boxes/regions of the image that very likely correspond to a bottom or a top. For the creation of bounding boxes/regions task, a benchmark dataset was specifically prepared. The results show that the Mask RCNN solution is robust, and generalized enough to be used and scalable in unseen apparel/fashion data.

cs.CV↗

arXiv2019

Combined collider constraints on neutralinos and charginos

Searches for supersymmetric electroweakinos have entered a crucial phase, as the integrated luminosity of the Large Hadron Collider is now high enough to compensate for their weak production cross-sections. Working in a framework where the neutralinos and charginos are the only light sparticles in the Minimal Supersymmetric Standard Model, we use gambit to perform a detailed likelihood analysis of the electroweakino sector. We focus on the impacts of recent ATLAS and CMS searches with 36 fb$^{-1}$ of 13 TeV proton-proton collision data. We also include constraints from LEP and invisible decays of the $Z$ and Higgs bosons. Under the background-only hypothesis, we show that current LHC searches do not robustly exclude any range of neutralino or chargino masses. However, a pattern of excesses in several LHC analyses points towards a possible signal, with neutralino masses of $(m_{\tildeχ_1^0}, m_{\tildeχ_2^0}, m_{\tildeχ_3^0}, m_{\tildeχ_4^0})$ = (8-155, 103-260, 130-473, 219-502) GeV and chargino masses of $(m_{\tildeχ_1^{\pm}}, m_{\tildeχ_2^{\pm}})$ = (104-259, 224-507) GeV at the 95% confidence level. The lightest neutralino is mostly bino, with a possible modest Higgsino or wino component. We find that this excess has a combined local significance of $3.3σ$, subject to a number of cautions. If one includes LHC searches for charginos and neutralinos conducted with 8 TeV proton-proton collision data, the local significance is lowered to 2.9$σ$. We briefly consider the implications for dark matter, finding that the correct relic density can be obtained through the Higgs-funnel and $Z$-funnel mechanisms, even assuming that all other sparticles are decoupled. All samples, gambit input files and best-fit models from this study are available on Zenodo.

hep-ph↗