obg7nkarlelt

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

論文三：

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翻譯社 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. 實驗的結果和其優良性：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，不是很明白）

6. 將來的研究偏向1：An intriguing question for future work is whether this framework can be useful for object detection翻譯社 in addition to recognition.

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. 進獻：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, and (3) integration of data-driven bottom-up process and context-driven top-down process for human activity understanding.

4. 實驗的結果：Experimental evaluation shows the efficacy of the proposed system for analyzing multi-person interactions.

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

6. 未來的研究標的目的：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,不外這是不准確的英文用法，建議改用Recommendations for future research)

Concluding remarks

Discussion

  貢獻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，然後用實驗結果去佐證。）

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

2.4 方法的限制2：However, the number of training samples required to directly represent the distribution of face images under varying pose may be prohibitively large.

6. 將來的研究標的目的3：Extrapolation in pose, e.g.翻譯社 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.

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，看來不太完備，花了一點時間，整理了一些材料，就用這一篇文章跟各人陳說天成翻譯公司的闡發成果和建議翻譯

結論的成份包孕下列幾項：

主軸一：

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. 英文的句型每每利用目前完成式，也有人利用過去式。

論文五：

1. 用一句話總結本篇論文提出的系統、方法或發現；

2. 系統的主要作法或方法的重點、應用、優點、和限制；

3. 進獻；（我很少加這一部份）

4. 嘗試的成績和其優良性；

5. 今朝系統的限制；

6. 將來的研究方向翻譯

Conclusion（我建議如果沒有特別的需求，就用這個；一般的論文，這應該是用得最多的。）

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.

論文四：

論文一：

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. 方式的優點：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. 方式的優點：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. 實驗的效果和其優良性：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, and numerical stability.

這篇論文的結論重點以方式的長處、限制等導到將來的研究建議，這裡的將來研究建議還分成三個來主軸討論翻譯

看似沒有完全頭緒的器械，顛末這五篇論文的分解，我們可以歸納成最前面的六個成分，這就是我的題目後半句話了：萬變不離其宗，你要著重在哪些點上，建議你和你的指點傳授討論一下。

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

本篇論文的貢獻寫在實行成果的後面，從邏輯上看也很准確；另請注意，本文沒有將來的研究建議。

結論的寫法存在良多的差別性，下面要解析的這幾篇文章的結論差異也很大。前面這兩篇論文結論的結構相對的純真，也很常見，建議大師參考利用。