图像生成与街景分割技术研究参与过程的学习经验总结

摘 要：我是一名计算机科学与技术专业的本科生。大二结束后的暑假，我第一次比较完整地进入课题组，开始接触图像风格迁移和街景语义分割两个方向。此前在课程学习中，我已经接触过 Transformer、注意力机制和语义分割，但更多时候只是知道概念、会看一些示例。真正读论文、跑模型之后，我才发现，知道一个概念和能够把它用起来，中间还有很长一段距离。过去一年多，我先从补基础开始，随后逐渐把注意力放到模型结构、实验过程和结果分析上，最后参与到技术材料的整理与写作中。本文按照这一过程展开，记录自己在知识理解、实验习惯以及科研思考方面的变化，也总结第一次参与科研训练后得到的一些体会。

关键词：学习经验；图像风格迁移；图像分割；本科生科研；深度学习实践；StyTR-2；DeeplabV3

正文 

可下载并阅读全文PDF，请按照本文版权许可使用。
Download the full text PDF for viewing and using it according to the license of this paper.

What I Learned from Taking Part in Image Generation and Street-Scene Segmentation Research

Abstract: I am an undergraduate student majoring in Computer Science and Technology. After the summer of my second year, I joined a research group and began working on two computer vision projects, involving Transformer-based image style transfer and street-scene semantic segmentation. I had encountered attention mechanisms, semantic segmentation, and convolutional neural networks in my coursework, but my understanding at that time was still largely based on definitions and small examples. The first time I had to read a complete research paper, run the corresponding code, and explain an unexpected experimental result, I realized that this knowledge was not yet enough for research work. During more than a year of training, I gradually worked through the full process from reviewing basic theory and understanding model structures to setting up environments, debugging, conducting experiments, organizing results, and preparing technical material. This paper records that process from the perspective of my own learning, with emphasis on several changes in the way I understood models, handled experiments, and approached research problems.

Keywords: Learning Experience; Image Style Transfer; Image Segmentation; Undergraduate Research; Deep Learning Practice; StyTR-2; DeeplabV3

参考文献

1. 陈淑環, 韦玉科, 徐乐, 等. 基于深度学习的图像风格迁移研究综述[J]. 计算机应用研究, 2019, 36(08): 2250-2255.
2. Chen L C, Papandreou G, Kokkinos I, et al. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 40(4): 834-848.
3. 陈淮源, 张广驰, 陈高, 等. 基于深度学习的图像风格迁移研究进展[J]. 计算机工程与应用, 2021, 57(11): 37-45.
4. 廉露, 田启川, 谭润, 等. 基于神经网络的图像风格迁移研究进展[J]. 计算机工程与应用, 2024, 60(09): 30-47.
5. Deng Y, Tang F, Dong W, et al. StyTR-2: Image Style Transfer with Transformers[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022.
6. Cordts M, Omran M, Ramos S, et al. The Cityscapes Dataset for Semantic Urban Scene Understanding[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 3213-3223.
7. Chollet F. Xception: Deep Learning with Depthwise Separable Convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017: 1251-1258.
8. Johnson J, Alahi A, Fei-Fei L. Perceptual Losses for Real-time Style Transfer and Super-resolution[C]//European Conference on Computer Vision (ECCV), 2016.
9. Park D Y, Lee K H. Arbitrary Style Transfer with Style-Attentional Networks[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
10. 张文海. 街景小目标识别算法研究[D]. 合肥: 中国科学技术大学, 2019.
11. 董康龙. 基于多尺度特征融合的街景图像分割算法研究[J]. 计算机应用, 2020, 40(5): 1345-1350.