零樣本學習是指分類器不只能識別在訓練階段已經看過的物件，甚至能識
別未曾看過的物件，而在多標籤零樣本學習中，每個實例中可能出現不只一個
物件，這使得識別任務變得更加困難。
　　過去的方法常利用標籤的屬性嵌入(attributes embedding)及影像抽取出的
視覺特徵(visual feature)，投影到同一空間中，藉此尋找與影像特徵最接近的
標籤，或是利用知識圖譜、知識庫建構標籤之間的關係，根據此關係來幫助辨
識標籤。然而在資料集欠缺屬性嵌入時，常用於替代的語義嵌入(word mbedding)並不像屬性嵌入一樣具有良好的辨識力，而建構關係的方法，也容易太過信任知識庫，便將關係強加上去，忽略了影像本身包含的資訊。近年來由於生成對抗網路(Generative Adversarial Network)的興起，對於未知類別，先從已知類別學習影像特徵的表達式及對應的屬性，再由屬性標籤生成影像特徵變得更加有效率，結果也更準確。基於這項觀察，我們提出了生成對抗網路結合語義嵌入的深度學習模型，從語義嵌入生成影像特徵，以及將影像特徵轉換成分類器映射至語義嵌入空間，尋找屬於該影像的標籤。藉由影像特徵及語義嵌入互相映射來更好地預測未知類別，並根據影像特徵與分類器之間的關係，將多標籤任務轉換化成單標籤任務。

[1] Wei Wang, Vincent W. Zheng, Han Yu, and Chunyan Miao. A Survey of Zero-Shot Learning: Settings, Methods, and Applications. ACM Trans. Intell. Syst. Technol.10, 2, Article 13, 2019.

[2] Andrea Frome, Greg S. Corrado, Jon Shlens, Samy Bengio, Jeff Dean, Marc'Aurelio Ranzato, Tomas Mikolov. DeViSE: A Deep Visual-Semantic Embedding Model. In NIPS, 2013.

[3] Elyor Kodirov, Tao Xiang, Shaogang Gong. Semantic Autoencoder for Zero-Shot Learning. In CVPR, 2017.

[4] Mehdi Mirza, Simon Osindero. Conditional Generative Adversarial Nets. arXiv preprint arXiv:1411.1784, 2014.

[5] Martin Arjovsky, Léon Bottou. Towards Principled Methods for Training Generative Adversarial Networks. In ICLR, 2017.

[6] Martin Arjovsky, Soumith Chintala, Léon Bottou. Wasserstein Generative Adversarial Networks. arXiv preprint arXiv:1701.07875, 2017.

[7] Xianwen Yu, Xiaoning Zhang, Yang Cao and Min Xia. VAEGAN: A Collaborative Filtering Framework based onAdversarial Variational Autoencoders. In IJCAI, 2019.

[8] Yongqin Xian, Sauabh Sharma, Bernt Schiele, Zeynep Akata. f-VAEGAN-D2: A Feature Generating Framework for Any-Shot Learning. In CVPR, 2019.

[9]Meng Ye, Yuhong Guo. Multi-Label Zero-Shot Learning with Transfer-AwareLabel Embedding Projection. arXiv preprint arXiv:1808.02474, 2018.

[10] Tomas Mikolov, Kai Chen, Greg Corrado, Jeffrey Dean. Efficient Estimation of Word Representations in Vector Space. In ICLR Workshop, 2013.

[11] Jeffrey Pennington, Richard Socher, Christopher D. Manning. GloVe: Global Vectors for Word Representation. In EMNLP, 2014.

[12] Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. arXiv preprint arXiv:1810.04805, 2018.

[13] M.L.Menéndez, J.A.Pardo, L.Pardo, M.C.Pardo. The Jensen-Shannon divergence. J. Frankl. Inst. 1997.

[14] M. Everingham, L. Gool, C. K. Williams, J. Winn, and A. Zisserman. The pascal visual object classes (voc) challenge. IJCV, 88(2):303–338, 2010.

[15] Tat-Seng Chua, Jinhui Tang, Richang Hong, Haojie Li, Zhip-ing Luo, and Yantao Zheng. Nus-wide: a real-world web image database from national university of singapore. In CIVR, 2009.

[16] Y. Zhang, B. Gong, and M. Shah, Fast zero-shot image tagging. In CVPR, 2016.

[17] M. B. Sariyildiz and R. G. Cinbis, Gradient matching generative networks for zero-shot learning. In CVPR, 2019.

[18] J. Lu, J. Li, Z. Yan, and C. Zhang, Zero-shot learning by generating pseudo feature representations. arXiv:1703.06389, 2017.

[19] Mohammad Norouzi, Tomas Mikolov, Samy Bengio, Yoram Singer, Jonathon Shlens, Andrea Frome, Greg S Corrado, and Jeffrey Dean. Zero-shot learning by convex combination of semantic embeddings. arXiv preprint arXiv:1312.5650, 2013.

[20] Y. Xian, Z. Akata, G. Sharma, Q. Nguyen, M. Hein, and B. Schiele. Latent embeddings for zero-shot classification. In CVPR, 2016.

[21] Y. Fu, Y. Yang, T. M. Hospedales, T. Xiang, and S. Gong. Transductive multi-label zero-shot learning. In BMVC, 2014.

[22] Zeynep Akata, Florent Perronnin, Zaid Harchaoui, and Cordelia Schmid. Label-embedding for image classification. In TPAMI, 2015.

[23] Dat Huynh and Ehsan Elhamifar. A shared multi-attention framework for multi-label zero-shot learning. In CVPR, 2020.

[24] Akshita Gupta, Sanath Narayan, Salman Khan, Fahad Shahbaz Khan, Ling Shao and Joost van de Weijer. Generative Multi-Label Zero-Shot Learning. arXiv preprint arXiv:2101.11606, 2021