Code and Data

Code

• MAGVIT (CVPR 2023) - The first vision transformer for multi-task video generation achieving state-of-the-art quality, efficiency, and flexibility.
• Generative Transfer Learning (CVPR 2023)- providing one-of-the-first benchmark results on generative transfer learning by transformers.
• MaskGIT (CVPR 2022) - The first Non-Autoregressive transformer for image synthesis
• LeCAM-GAN (CVPR 2021) - A simple and effective regularization approach to learn robust GAN on limited training data.
• MentorMix (ICML 2020)- A robust deep learning method for realistic noisy labels.
• Future Prediction in Video (CVPR 2019) - The first model for joint path and activity prediction.
• Text Image Residual Gating (CVPR 2019) - The first method for using vision and language modification for retrieval.
• Eidetic-3D LSTM (ICLR 2019)- State-of-the-art of self-supervised video learning model.
• MentorNet (ICML 2018) - Weakly-supervised deep learning method.
• Self-paced Learning (NeurIPS 2014) - An implementation for self-paced learning used in our paper.

Data Set

• Controlled Noisy Web Labels (ICML 2020) - First dataset and benchmark for realistic, real-world label noise sourced from the web.
• MemexQA (TPAMI 2019) - Multimodal dataset consisting of real personal photos and crowd-sourced questions/answers.
• YouTube-8M - Large-scale labeled video dataset consisting of millions of YouTube videos.
• Data for Image Manipulation by Text Instruction (ACM Multimedia 2020)- Dataset used for image manipulation by text instruction.
• CMU Viral Video Dataset (ICMR 2014) - Public dataset for viral video study.