About preprocess

[lzkzls]

Thanks for your sharing! There was a problem when I was reading the code. Do you need to generate "young / valfilenames. txt", "lrw.txt", “_mouth.txt”, and "_filenames.txt" before you running the process256_224.py? I downloaded the LRW dataset, which does not contain these files.

[Hangz-nju-cuhk]

Yes, although I almost forget the details of my preprocessing, the lrw.txt and filenames are basically the class list and filename list. The most important part is the _mouth.txt which contains the detected landmarks of faces. However, we cannot provide such files due to the license of the landmark detection tool. You can use any open-sourced landmark detection tool and modify the code. All we need is acutally 3 points, the eyes and the center of the mouth.

[love112358]

有谁知道_mouth.txt 这个文件的里面的数据是怎么处理得到的吗？里面包含的参数维度分别是什么样子的呢？谢谢

[love112358]

大家好，想跟大家请教一下关于mouth.txt 的数据处理过程，mouth.txt 的数据分别表示什么，多少维度的，我没有下载作者使用的数据集，我想使用新闻主持人的数据，但是我不知道如果生成mouth.txt 文件。如果有人处理处mouth.txt的数据，希望能告诉我mouth.txt的数据是多少维度的 分别表示什么意思，谢谢@liuziwei7 @liuyuisanai @Hangz-nju-cuhk @lzkzls 请问一下你知道如何处理mouth.txt的数据了吗