著急畢業(yè)投了一個mdpi旗下的水刊，投稿將近一個月收到了兩個審稿人意見，然后編輯就發(fā)了郵件pending major revision，給了10天時間要求大修（第三個審稿人還沒回復(fù)），第一次投稿沒啥經(jīng)驗，想請大家?guī)兔�參謀一下應(yīng)該按照什么方向來改呢？
第一個審稿人打分雖然相對比較高，但是他覺得創(chuàng)新性不夠建議以letter形式發(fā)表(我自己確實也覺得創(chuàng)新性不夠，但是創(chuàng)新點(diǎn)這個東西比較主觀，而且投稿的畢竟是水刊)；第二個審稿人雖然打分沒有第一個高，但問題都直擊要害，應(yīng)該是很懂行的專家，不過感覺認(rèn)真回答問題，改一改補(bǔ)充一點(diǎn)實驗還是比較好過的，沒有對文章的創(chuàng)新性提出質(zhì)疑。
問題就是我現(xiàn)在不知道是否應(yīng)該改成letter，畢竟這樣改動還是很大的，由于第二個審稿人的問題中需要有補(bǔ)充實驗的地方，怕時間來不及，而且改成letter的話會減少很多實驗，不知道第二個審稿人會不會不同意，但是不改letter的話又怕第一個審稿人直接拒，或者說有沒有什么比較好的方式回答第一個審稿人關(guān)于創(chuàng)新性的質(zhì)疑呢(同時看審稿意見我感覺第一個審稿人沒有特別仔細(xì)讀我的文章，有些地方理解的跟我想表達(dá)的甚至跟第二個審稿人理解的明顯不一樣)…想請大家?guī)兔�看看這種情況應(yīng)該怎么改呢？附圖分別是第一個和第二個審稿人的打分
下面是兩個審稿人的總體意見：
審稿人一：
this paper proposed a deep learning based demosaicking method using several existing techniques, such as the unet++ framework, residual in residual dense block, and depthwise separable convolutions. in particular, the framework inserts densely connected layer blocks that adopt depthwise separable convolutions to reduce the number of parameters. the main novelty of the paper is that: deploying the gaussian smoothing layer into the cnn framework can expand the receptive field without down-sampling image size, achieving the fastest execution time and modest quality performance relative to the considered demosaicking methods.??because the novelty in this paper is limited, we recommend that this paper should be revised and shortened for possible publication as a `letter'.
審稿人二：
the authors present a method for fast and efficient demosaicking. the manuscript contains some interesting ideas and experiments comparing the performance of their method to other methods. in general it is well written and easy to read. also the length is sufficient for the topic. my three biggest concerns on the manuscripts are:?
1) by making multiple numbers boldface the focus is taken away from the best performing methods in each of the tables.
2) drawing run-time conclusions from comparing cpu algorithms to gpu algorithms is unfair and favors the gpu versions.
3) the comparison of the cascade: demosaick->classification and just the classification on the original seems to favor the demosaick->classification pipeline because images are greatly reduced in size anyway.
具體的意見就不放了，兩個審稿人加起來提了13條左右的問題，包括讓補(bǔ)充指定文獻(xiàn)和書寫錯誤之類的

 返回小木蟲查看更多