PinnedInTDS ArchivebyMichael Sugimura·May 20, 2019FGO StyleGAN: This Heroic Spirit Doesn’t ExistFinetuning Nvidia’s StyleGAN to a custom Fate Grand Order use caseA response icon6A response icon6
Michael Sugimura·Mar 19, 2021A Tricksy Look at Outfitting: KitsuneAn approach to outfitting using attribute vector representations and a multi-task generator model to create outfits for a given seed item
InTDS ArchivebyMichael Sugimura·May 11, 2020Project Pendragon + Tonks: Multi-Task Feature Extraction for Farming Fate Grand OrderUsing a Tonks trained multi-task pytorch model to extract features from the mobile phone game Fate Grand Order
Michael Sugimura·Apr 28, 2020Tonks: Building One (Multi-Task) Model to Rule Them All!a new library for training multi-task networks in pytorch with images and or text.A response icon1A response icon1
InTDS ArchivebyMichael Sugimura·Feb 25, 2020Breaking the Game: Pendragon Four Rise of MerlinMulti-agent reinforcement learning for the mobile phone game Fate Grand Order adding in powerful supports to clear late-game content
InTDS ArchivebyMichael Sugimura·Feb 17, 2020Pendragon Four: Training Pipeline Deeper Dive for Multi Agent Reinforcement LearningDeeper dive into training multiple RL agents simultaneously to play the mobile phone game Fate Grand Order
InTDS ArchivebyMichael Sugimura·Dec 17, 2019Pendragon Four: Multi-Agent Reinforcement Learning with Fate Grand OrderMulti-agent reinforcement learning in a custom game environment to train 4 agents and have them play the mobile phone game Fate Grand Order
InTDS ArchivebyMichael Sugimura·Oct 9, 2019This Dress Doesn’t ExistFine tuning GPT-2 and StyleGAN for a fashion use case to generate synthetic product images and descriptionsA response icon2A response icon2
InTDS ArchivebyMichael Sugimura·Sep 9, 2019Building and Labeling Image Datasets for Data Science ProjectsSome tips and tricks for building image datasetsA response icon2A response icon2
InTDS ArchivebyMichael Sugimura·Jun 10, 2019BERT Classifier: Just Another Pytorch ModelBuilding a custom Pytorch pipeline for a BERT classifier to figure out how BERT works piece by pieceA response icon3A response icon3