Mahyar Gohari

Research

I'm interested in computer vision, signal processing, and machine learning.

Audio Features Investigation for Singing Voice Deepfake Detection

This research investigates the effectiveness of various audio representations and features for discriminating real and synthetically generated singing voice signals.

Spectrogram-Based Detection of Auto-Tuned Vocals in Music Recordings

This study introduces a data-driven approach leveraging triplet networks for the detection of Auto-Tuned songs.

PGNN-based Approach for Robust 3D Light Direction Estimation in Outdoor Images

This research introduces a novel methodology utilizing a physics-guided neural network (PGNN) to achieve precise global 3D light direction estimation. The proposed architecture integrates an illumination model, allowing the network to indirectly acquire geometric information, thereby enhancing the accuracy of the estimated light direction.

SynthOutdoor: a synthetic dataset for 3D outdoor light estimation

In this work, we present SynthOutdoor dataset, comprising 39086 high-resolution images, aimed at addressing the data scarcity in the field of 3D light direction estimation under the assumption of distant lighting. SynthOutdoor was generated using our software that is based on the Unity3D engine. Our dataset provides a set of images rendered from a given input scene, with the camera moving across a predefined path within the scene. This dataset captures a wide variety of lighting conditions through the implementation of a solar cycle.

Detection and Localization of Ripe Tomatoes Using Machine Vision

The study aimed to design a model for a tomato harvester robot using machine vision to be able to detect and locate tomatoes automatically and in real-time.

Source code from Leonid Keselman.