Snake emoji的問題，透過圖書和論文來找解法和答案更準確安心。 我們找到下列訂位、菜單、價格優惠和問答集

基於 TraVeLGAN 與 Perceptual Loss 實現照片轉換表情符號之應用

為了解決Snake emoji的問題，作者黃銘俊 這樣論述：

Cartoon is one of the media that could both convey information and induce entertainment effects. Cartoonization of real photos is therefore of interest, and may be achieved through image transformation approaches such as generative adversarial networks. However, conventional generative adversarial

network may suffer from mode collapse, i.e. generating highly similar images. This is because the only constraint imposed on the generator is to produce something similar to real images, so it leads to multiple-to-one mapping. TraVeLGAN, on the other hand, is the network that may tackle this issue b

y forcing the synthesized images separated and keeping their spatial relationship similar to that of the original images. In this study, image transformation for human face from photo-realistic image to emoji image based on TraVeLGAN is investigated. In the initial experiment, TraVeLGAN may generate

the images with higher diversity, but they often have mismatched semantic attributes, such as hair and skin color, or shape of the head. To alleviate this problem, perceptual loss computed from VGG19 is proposed to be used with TraVeLGAN, since perceptual loss may make the output image closer to th

e input image on the feature map. Experimental result shows TraVeLGAN can produce the images with better quality and higher SSIM score. In addition, perceptual loss obtained from a shallower layer, such as the first or second convolutional layer, may give higher similarity and better quality. Furthe

rmore, a generative adversarial network with perceptual loss is conducted for comparison, and it is found that TraVeLGAN is helpful for improving image quality.Keywords: cartoonization, photo-to-emoji transformation, Siamese network, generative adversarial network, TraVeLGAN, perceptual loss. 

非言語性線索在即時通訊軟體中所扮演的角色——以貼圖為例

為了解決Snake emoji的問題，作者林堂智 這樣論述：

處在智慧型裝置盛行的時代，人們不再需要面對面就能傳遞訊息。但在非面對面的人際溝通過程中，非語言性訊息的缺乏容易導致人們誤解訊息內容。在實驗一，主要探討非言語性線索是否能夠做到傳遞情緒的功能，故操弄了人際關係（好朋友、陌生人）與貼圖情境（正向貼圖、負向貼圖、模棱兩可貼圖）。結果發現，人際關係會影響到人們對貼圖的情緒感知，但是在負向貼圖情境則無人際關係間差異。在選擇回覆的貼圖上，不論發訊者的人際關係，實驗參與者都會回覆與其情緒向度一致的貼圖，唯有在與好朋友的情境中，實驗參與者較會使用模棱兩可的貼圖。實驗二檢視當只有言語性線索時（正向文字、負向文字），貼圖的選擇是否會受到情境以及人際關係的調節。情

緒感知結果顯示，若是好朋友發送正向文字時，接收端感知的情緒更為正向，而陌生人發送負向文字則會被感知為更為負向。在貼圖回覆結果發現，不論發送端的人際關係，實驗參與者都會回覆與其情緒向度一致的貼圖。實驗三 結合文字與貼圖情境，結果發現，情緒感知的結果與實驗一相似，唯有在搭配負向貼圖時，人際關係並無情緒感知上的差異。當文字情境與貼圖情境相左時，實驗參與者傾向做出與貼圖情境一致的回覆；文字情境與貼圖情境一致時，實驗參與者傾向給予一致的回覆。但是，在模棱兩可的貼圖回覆上，當發訊者為陌生人，且文字情境與貼圖情境相左時，實驗參與者選擇模棱兩可貼圖的比例較其他貼圖來得高。綜合三個實驗的結果，本研究發現不同人際

關係的發送端會影響人們對於正向情緒的感知，但負向情緒的感知則不會任何的差異。文字固然能夠傳遞情緒訊息，但貼圖卻能夠影響人們對文字訊息內容的情緒感知，因此錯誤使用貼圖將會形成一種干擾。此外，若要正確無誤的使用即時軟體傳達訊息，情緒貼圖的選擇是重要的，其會大大地影響到人們對訊息情緒的詮釋，以及作出怎樣的回覆。