吃面的重点问题我认为有两个，一：被夹起的面；二：嘴里的面。对这些图片的tag描述不够细致，以至于容易出现用手抓面或是块状的面出现在嘴边。我用了 noodles_in_mouth, clutch_noodles 两个tag来描述这两个方面，但仍有一些问题——1：会出现三根筷子（也许训练 3chopsticks 这样的负面词条会改善这个问题）；2：会在图片的随机位置出现被夹起的面（需要分层定义tag的位置 我应该会在下个版本迭代掉这个问题）

（clutch_noodles）

（noodles_in_mouth）

（Noodles appeared where I didn't want it to be）

I think there are two key issues in eating noodles: one: the noodles being picked up; and two: the noodles in the mouth. The tag descriptions of these images are not detailed enough, so that it is easy to grab the noodles by hand or to have chunks of noodles appearing in the mouth. I used the tags noodles_in_mouth, clutch_noodles to describe these two aspects, but there are still some problems - 1: three chopsticks appear (perhaps training negative entries like 3chopsticks would improve this problem); 2: noodles appear in random locations in the picture random positions (need to define tag positions in layers I should iterate out this problem in the next version)