willard-yuan / video-text-retrieval-papers

Awesome Text to Video papers

The main goal is to collect text-to-video works in academia and industry.

Papers

• Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision, arxiv 2021.
• Frozen in Time: A Joint Video and Image Encoder for End-to-End Retrieval, arxiv 2021, code.
• MDMMT: Multidomain Multimodal Transformer for Video Retrieval, arxiv 2021.
• Multi-modal Transformer for Video Retrieval, ECCV 2020, code.
• VisRel: Media Search at Scale, SIGKDD 2021.
• Que2Search: Fast and Accurate Query and Document Understanding for Search at Facebook, SIGKDD 2021.
• CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retrieval, code.

Datasets

• Microsoft Research Video Description Corpus (MSVD)：也称为YouTube2Text dataset，该数据集同样由Microsoft Research提供，地址为 Microsoft Research Video Description Corpus 。该数据集包含1970段YouTube视频片段（时长在10-25s之间），每段视频被标注了大概40条英文句子。
• MSR-VTT (Microsoft Research Video to Text)：该数据集为ACM Multimedia 2016 的 Microsoft Research - Video to Text (MSR-VTT) Challenge。地址为 Microsoft Multimedia Challenge 。该数据集包含10000个视频片段（video clip），被分为训练，验证和测试集三部分。每个视频片段都被标注了大概20条英文句子。此外，MSR-VTT还提供了每个视频的类别信息（共计20类），这个类别信息算是先验的，在测试集中也是已知的。同时，视频都是包含音频信息的。该数据库共计使用了四种机器翻译的评价指标，分别为：METEOR, BLEU@1-4,ROUGE-L,CIDEr。
• LSMDC (Large Scale Movie Description Challenge): This dataset contains 118,081 short video clips extracted from 202 movies. Each video has a caption, either extracted from the movie script or from transcribed DVS (descriptive video services) for the visually impaired. The validation set contains 7408 clips and evaluation is performed on a test set of 1000 videos from movies disjoint from the training and val sets.

Benchmarks

• Text-To-Video Benchmarks.