arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1806.04768cs.CV
收藏

无监督的分层长期视频预测

Hierarchical Long-term Video Prediction without Supervision

Google Brain(谷歌大脑);University of Michigan(密歇根大学)

Nevan Wichers, Ruben Villegas, Dumitru Erhan, Honglak Lee

摘要:

最近的大部分研究都致力于视频预测和生成,但大多数以前的作品在短期内生成视频方面只取得了有限的成功。Villegas等人(2017)的分层视频预测方法是用于长期视频预测的最先进方法的一个示例,但他们的方法是有限的,因为它需要高层结构的地面实况注释(例如,人关节界标)。我们的网络对输入帧进行编码,预测未来的高级编码,然后访问第一帧的解码器从预测编码中生成预测图像。解码器还产生描绘预测前景对象的轮廓的掩模(例如,人)作为副产品。与Villegas et al.(2017)不同,我们开发了一种新的训练方法,可以在没有高级监督的情况下联合训练编码器,预测器和解码器;我们通过在特征空间中使用对抗性损失来训练预测器来进一步改进。我们的方法可以预测未来大约20秒,并且与Denton和Fergus(2018)和Finn等人(2016)在人类3.6M数据集上相比,提供了更好的结果。

英文摘要:

Much of recent research has been devoted to video prediction and generation, yet most of the previous works have demonstrated only limited success in generating videos on short-term horizons. The hierarchical video prediction method by Villegas et al. (2017) is an example of a state-of-the-art method for long-term video prediction, but their method is limited because it requires ground truth annotation of high-level structures (e.g., human joint landmarks) at training time. Our network encodes the input frame, predicts a high-level encoding into the future, and then a decoder with access to the first frame produces the predicted image from the predicted encoding. The decoder also produces a mask that outlines the predicted foreground object (e.g., person) as a by-product. Unlike Villegas et al. (2017), we develop a novel training method that jointly trains the encoder, the predictor, and the decoder together without highlevel supervision; we further improve upon this by using an adversarial loss in the feature space to train the predictor. Our method can predict about 20 seconds into the future and provides better results compared to Denton and Fergus (2018) and Finn et al. (2016) on the Human 3.6M dataset.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑