今年,我很開心能與同事、教授與學生們合力執行一些新嘗試,並發表於學術會議上。讓我帶著一點驕傲、一點幽默,還有一點點智慧 (盡力),來分享這些成就吧:
1️⃣ PiCoGen: Generate Piano Covers with a Two-stage Approach
「為什麼要上網找琴譜,為什麼要自己彈鋼琴改編曲,當 AI 就可以幫你搞定的話?」PiCoGen 不需要傳統繁瑣的配對資料集訓練,就能直接從一首流行歌曲中提取出琴譜來,並且直接彈奏給你聽。鋼琴家們,小心了 XD
2️⃣ SARA: Semantic-Assisted Reinforced Active Learning for Entity Alignment
你家的知識圖譜不聽話?讓 SARA 來幫忙!這是主動學習結合語義的研究,專治資料少、異構性強的後設資料集。就像一位耐心的調解專家,讓圖譜們終於願意「攜手合作」。
3️⃣ Description-Driven Audiovisual Embedding Space Learning for Enhanced Movie Understanding
電影 + AI 很厲害?當然要自行試試看才知曉!從自動下標籤到推測電影類型,這篇研究深入影音的裏世界。終於,我們有一個演算法可以比那位每次看片都睡著的朋友更“稍微的“懂電影了~
4️⃣ PiCoGen2: Piano Cover Generation with Transfer Learning Approach and Weakly Aligned Data
如果說 PiCoGen 是產出是一首旋律,那 PiCoGen2 帶來的就是精彩的改編版。透過遷移學習與巧妙的資料處理方式,這個模型即便在資料有限的情況下,仍能創造出更高質量的鋼琴改編。莫札特可能會嫉妒了 (會嗎 XD)。
- Chih-Pin Tan, Shuen-Huei Guan, and Yi-Hsuan Yang, “PiCoGen: Generate Piano Covers with a Two-stage Approach,” ACM International Conference on Multimedia Retrieval (ICMR), 2024.
- Ching-Hsuan Liu, Chih-Ming Chen, Jing-Kai Lou, Ming-Feng Tsai, Jiun-Lang Huang, and Chuan-Ju Wang, “SARA: Semantic-assisted Reinforced Active Learning for Entity Alignment,” IEEE International Joint Conference on Neural Networks (IJCNN), 2024.
- Wei-Lun Huang, Shao-Hung Wu, Hung-Chang Huang, Min-Chun Hu, and Tse-Yu Pan, “Description-Driven Audiovisual Embedding Space Learning for Enhanced Movie Understanding,” ACM International Conference on Multimedia in Asia (MMAsia), 2024.
- Chih-Pin Tan, Hsin Ai, Yi-Hsin Chang, Shuen-Huei Guan, and YI-Hsuan Yang, “PiCoGen2: Piano Cover Generation with Transfer Learning Approach and Weakly Aligned Data,” International Society for Music Information Retrieval Conference (ISMIR), 2024.