{"title":"帧插值使用深金字塔流","authors":"Shangshu Qian, Zhi-he Zhou, Shuyue Lai","doi":"10.1109/GEOINFORMATICS.2018.8557169","DOIUrl":null,"url":null,"abstract":"In this paper, we made an attempt to generate the middle frame by analyzing an existing video. While the blurry intermediate frames generated by current methods, which are widely used, are not quite satisfying, our method combines the advantages of estimating optical flow and hallucinating the RGB value directly, which is able to generate sharp and realistic frames. Apart from that, it is unsupervised, and SSIM is applied as loss function to better our outcome. Eventually, our method was tested on the UCF-101 and THUMOS-15 dataset. It turned out that the above-mentioned approach produces high-quality and visually acceptable results, which outperforms other competing methods.","PeriodicalId":142380,"journal":{"name":"2018 26th International Conference on Geoinformatics","volume":"23 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Frame Interpolation Using Deep Pyramid Flow\",\"authors\":\"Shangshu Qian, Zhi-he Zhou, Shuyue Lai\",\"doi\":\"10.1109/GEOINFORMATICS.2018.8557169\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In this paper, we made an attempt to generate the middle frame by analyzing an existing video. While the blurry intermediate frames generated by current methods, which are widely used, are not quite satisfying, our method combines the advantages of estimating optical flow and hallucinating the RGB value directly, which is able to generate sharp and realistic frames. Apart from that, it is unsupervised, and SSIM is applied as loss function to better our outcome. Eventually, our method was tested on the UCF-101 and THUMOS-15 dataset. It turned out that the above-mentioned approach produces high-quality and visually acceptable results, which outperforms other competing methods.\",\"PeriodicalId\":142380,\"journal\":{\"name\":\"2018 26th International Conference on Geoinformatics\",\"volume\":\"23 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2018-06-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2018 26th International Conference on Geoinformatics\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/GEOINFORMATICS.2018.8557169\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 26th International Conference on Geoinformatics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/GEOINFORMATICS.2018.8557169","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
In this paper, we made an attempt to generate the middle frame by analyzing an existing video. While the blurry intermediate frames generated by current methods, which are widely used, are not quite satisfying, our method combines the advantages of estimating optical flow and hallucinating the RGB value directly, which is able to generate sharp and realistic frames. Apart from that, it is unsupervised, and SSIM is applied as loss function to better our outcome. Eventually, our method was tested on the UCF-101 and THUMOS-15 dataset. It turned out that the above-mentioned approach produces high-quality and visually acceptable results, which outperforms other competing methods.