christdaycibj

英翻塞爾維亞文

主軸二：

論文二：

結論的寫法存在良多的差別性，下面要剖析的這幾篇文章的結論差別也很大。前面這兩篇論文結論的佈局相對的純真，也很常見，建議大家參考應用。 

  進獻2：2) Gabor real and imaginary parts are successfully combined with LDP and LBP. Experimental results show that both LDP and LBP on Gabor feature images achieve much better performance than LDP and LBP on gray-level images.（如上，進獻用一句話作申明，然後用嘗試效果去左證。）

1. 用一句話總結本篇論文提出的方式：This paper investigates the feasibility and effectiveness of using high-order local pattern for face description and recognition. （用This paper開頭時可以使用而今式，因為這是事實。）

2.1. 方式的重點：A Local Derivative Pattern (LDP) is proposed to capture the high-order local derivative variations. To model the distribution of LDP micropatterns, an ensemble of spatial histograms is extracted as the representation of the input face image.

2.2. 方式的運用：Face recognition based on LDP can be performed by using histogram intersection as the similarity measurement.

4. 實行的成績和其優良性：Experimental results on an extensive set of face databases, FERET, CAS-PEAL, CMU-PIE翻譯社 Extended Yale B翻譯社 and FRGC databases翻譯社 demonstrate that the proposed high-order local pattern representation outperforms LBP representation in both identification and verification.

3. 貢獻1：The main contributions of this paper include: 1) A novel local descriptor, high-order Local Derivative Pattern翻譯社 is proposed as an object descriptor. Experiments conducted on various face conditions, including different lightings, expressions, agings, accessories and poses, show that high-order local patterns (LDPs) achieve better performances than the first-order local pattern (LBP). （貢獻用一句話作申明，然後用實行結果去佐證。）

主軸一：

2. 系統的首要作法：The analysis is performed in a distributed vision system (called ‘Track-Body Synergy’ (TBS)) that synergistically integrates track- and body-level representations from multiple views.

3. 貢獻：The main contributions of the paper are: (1) context-dependent view switching for occlusion handling翻譯社 (2) switching the multi-level analysis between track- and body-level representations翻譯社 and (3) integration of data-driven bottom-up process and context-driven top-down process for human activity understanding.

4. 嘗試的效果：Experimental evaluation shows the efficacy of the proposed system for analyzing multi-person interactions.

5. 今朝系統的限制：Our current implementation uses two cameras翻譯社 but the extension to more cameras is straightforward.

6. 將來的研究偏向：We plan in the future to implement multi-thread processes for diverse selections of the views. The use of a hierarchical HMM structure for recognizing more general activity patterns also belongs to our future plan. （在學位論文中有人會加上題目：Future research directions,不外這是不正確的英文用法，建議改用Recommendations for future research)

翻譯公司選用了哪一個，重點要隨著調劑。

結論的成份包括以下幾項：

論文四：

2.3. 方式的長處2：Robustness to occlusion allows the algorithm to tolerate small pose variation or misalignment.

2.3. 方式的長處3：Furthermore翻譯社 in the supplementary appendix翻譯社 which can be found on the Computer Society Digital Library at http://doi.ieeecomputersociety.org/ 10.1109/TPAMI.2008.79, we discuss our algorithm’s ability to adapt to nonlinear training distributions.

2.4 方式的限制2：However翻譯社 the number of training samples required to directly represent the distribution of face images under varying pose may be prohibitively large.

6. 未來的研究偏向3：Extrapolation in pose, e.g., using only frontal training images, will require integrating feature matching techniques or nonlinear deformation models into the computation of the sparse representation of the test image.

這篇論文的結論重點以方式的優點、限制等導到未來的研究建議，這裡的將來研究建議還分成三個來主軸討論。

5. Conclusions and future work

1. 用一句話總結本篇論文提出的方式：A solution to enhance the performance of classical mean shift object tracking has been presented. （而今完成式）

2. 方法的重點：This work integrated the outcomes of SIFT feature correspondence and mean shift tracking. An expectation–maximization algorithm was proposed to optimize the probability function for a better similarity search. （用過去式，天成翻譯公司建議採用！）

4. 實驗的成績和其優良性：Experiments verified that the proposed method could produce better solutions in object tracking of different scenarios. （也是曩昔式）

6. 將來的研究方向：In future work, the research attempt is to investigate the convergence roperty of the proposed framework. This investigation may help enhance the proposed algorithm for efficiency purposes. In addition, this proposed algorithm needs to be comprehensively evaluated in a wider database. Currently, this paper suggests that, although the tracking results are promising in certain situations, further development and more evaluation is anticipated in severe image clutters and occlusions.（這個部份相對的比力長；還有，留意，動詞的時態利用而今式翻譯）

Conclusion and discussion

Discussion

論文五：

本篇論文的結論沒有將來的研究建議。

1. 用一句話總結本篇論文提出的系統、方式或發現；

2. 系統的首要作法或方式的重點、利用、長處、和限制；

3. 貢獻；（我很少加這一部份）

4. 嘗試的後果和其優良性；

5. 今朝系統的限制；

6. 將來的研究標的目的。

Conclusion（我建議若是沒有稀奇的需求，就用這個；一般的論文，這應該是用得最多的翻譯）

Concluding remarks

本篇論文的進獻寫在實行結果的後面，從邏輯上看也很正確；另請留意，本文沒有將來的研究建議。

2.4 方法的限制1：The usefulness of sparsity in detection has been noticed in the work in [61] and more recently explored in [62]. We believe that the full potential of sparsity in robust object detection and recognition together is yet to be uncovered.

6. 將來的研究偏向2：From a practical standpoint, it would also be useful to extend the algorithm to less constrained conditions翻譯社 especially variations in object pose.

論文三：

Concluding remarks
1. 用一句話總結本篇論文提出的系統：We have presented a two-stage multi-view analysis framework for understanding human activity and interactions. 英文的句型凡是利用此刻完成式，也有人利用曩昔式。

1. 用一句話總結本篇論文提出的方式1：In this paper, we proposed a new method for addressing computational difficulties encountered in obtaining the optimal projection vectors in the null space of the withinclass scatter.

2.3. 方式的長處：We showed that every sample in a given class produces the same unique common vector when they are projected onto the null space of SW.

1. 再用一句話總結本篇論文提出的方式2：We also proposed an alternative algorithm for obtaining common vectors based on the subspace methods and the Gram-Schmidt orthogonalization procedure, which avoids handling large matrices and improves the stability of the computation.

2.3. 方式的長處：Using common vectors also leads to an increased computational efficiency in face recognition tasks.

2.1. 方式的重點：The optimal projection vectors are found by using the common vectors and the discriminative common vectors are determined by projecting any sample from each class onto the span of optimal projection vectors.

2.3. 方式的長處：There is no loss of information content in our method, in the sense that the method has 100 percent recognition rate for the training set data.

4. 實行的了局和其優良性：Experimental results show that our method is superior to other methods in terms of accuracy翻譯社 real-time performance翻譯社 storage requirements翻譯社 and numerical stability.

進入例子理會前，先看一下結論的標題；英文學術論文中結論的用詞其實很不相同：

論文一：

2.1. 方式的重點/發現：With sparsity properly harnessed, the choice of features becomes less important than the number of features used (in our face recognition example, approximately 100 are sufficient to make the difference negligible). Moreover, occlusion and corruption can be handled uniformly and robustly within the same classification framework.

4. 實行的結果和其優良性：One can achieve a striking recognition performance for severely occluded or corrupted images by a simple algorithm with no special engineering. （用one當主詞，不是很明確）

6. 將來的研究標的目的1：An intriguing question for future work is whether this framework can be useful for object detection, in addition to recognition.

主軸三：

針對學術論文的寫法，天成翻譯公司前面寫了摘要和導論兩大部份，不來介紹一下結論的寫法，看來不太完整，花了一點時間，整理了一些材料，就用這一篇文章跟大家陳訴我的闡明效果和建議翻譯

看似沒有完全頭緒的東西，經由這五篇論文的理會，我們可以歸納成最前面的六個成份，這就是我的題目後半句話了：萬變不離其宗，你要側重在哪些點上，建議你和你的指點傳授接洽一下翻譯

Doing so, in a principled manner, it remains an important direction for future work.

Concusions and future work

1. 用一句話總結本篇論文提出的方式：In this paper, we have contended both theoretically and experimentally that exploiting sparsity is critical for the high-performance classification of high-dimensional data such as face images.