🚀 [2024.01.19] New Blog Post: ESPnet Hackathon 2023✨

https://anthony-wss.github.io/

• The experiments of CLCIFAR, the first human-annotated complementary label datasets.

• Collect 30k human annotated complementary labels for CIFAR10 and CIFAR100 with Amazon Mechanical Turk.

• Design batch parallel pipeline to identify the language id of 200k hours audio data scrapped from Youtube with a pre-trained deep learning model.

• Design and manage a robust server-client socket structure on the workstation

• Rank 10, 3, 6 out of 40 teams for naturalness, relevance, and aggressiveness by fine-tuning a T5-based chatbot inferenced with beam search developed by my teammates.