返回
SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks

SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks

ID：40725313

大小：207.07 KB

页数：5页

时间：2019-08-06

SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks _第1页
SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks _第2页
SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks _第3页
SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks _第4页
SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks _第5页
资源描述：

《SAD-GAN -- Synthetic Autonomous Driving using Generative Adversarial Networks 》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、SAD-GAN:SyntheticAutonomousDrivingusingGenerativeAdversarialNetworksArnaGhosh∗BiswarupBhattacharya∗DepartmentofElectricalEngineeringDepartmentofElectricalEngineeringIndianInstituteofTechnologyIndianInstituteofTechnologyKharagpur,WB721302.India.Kharagpur,WB721302.India.arnaghosh@iitkgp.ac.inbiswaru

2、p@iitkgp.ac.inSomnathBasuRoyChowdhury∗DepartmentofElectricalEngineeringIndianInstituteofTechnologyKharagpur,WB721302.India.brcsomnath@ee.iitkgp.ernet.inAbstractAutonomousdrivingisoneofthemostrecenttopicsofinterestwhichisaimedatreplicatinghumandrivingbehaviorkeepinginmindthesafetyissues.Weapproacht

3、heproblemoflearningsyntheticdrivingusinggenerativeneuralnetworks.Themainideaistomakeacontrollertrainernetworkusingimagespluskeypressdatatomimichumanlearning.WeusedthearchitectureofastableGANtomakepredictionsbetweendrivingscenesusingkeypresses.Wetrainourmodelononevideogame(RoadRash)andtestedtheaccu

4、racyandcompareditbyrunningthemodelonothermapsinRoadRashtodeterminetheextentoflearning.1IntroductionSelf-drivingcarsareoneofthemostpromisingprospectsforneartermartificialintelligenceresearch.Autonomousdrivingisawell-establishedproblemandtheuseoflargeamountsoflabeledandcontextuallyrichdatatosolvethep

5、roblemsofroaddetectionandpredictionofvehicleparameterslikeaccelerator,clutchandbrakepositionshavealreadybeenexplored[5].However,amajorchallengeisadatasetthatissufficientlyrichtocoverallsituationsaswellasdifferentconditions.arXiv:1611.08788v1[cs.CV]27Nov2016Asolutionproposedtoaidtheissueistheuseofsy

6、ntheticdataalongwithnaturaldatatotrainthesystem[12].Drivingisataskthatdemandscomplicatedperceptionandcontrolstaskswhichareintricatelylinkedtoeachother.Thetechnologytocorrectlysolvedrivingcanpotentiallybeextendedtootherinterestingtaskssuchasactionrecognitionfromvideosandpathplanninginrobotics.Visio

7、nbasedcontrolsandreinforcementlearninghadrecentsuccessintheliterature[6],[9],[14],[8]mostlydueto(deep,recurrent)neuralnetworksandunboundedaccesstoworldorgameinteraction.Suchinteractionsprovidethepossibilitytorevi

当前文档最多预览五页，下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页，下载文档查看全文
温馨提示：
1. 部分包含数学公式或PPT动画的文件，查看预览时可能会显示错乱或异常，文件下载后无此问题，请放心下载。
2. 本文档由用户上传，版权归属用户，天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容，确认文档内容符合您的需求后进行下载，若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误，付费完成后未能成功下载的用户请联系客服处理。