Finish-to-end Generative Pre-training for Multimodal Video Captioning

0
17


Multimodal video captioning techniques make the most of each the video frames and speech to generate pure language descriptions (captions) of movies. Such techniques are stepping stones in the direction of the longstanding purpose of constructing multimodal conversational techniques that effortlessly talk with customers whereas perceiving environments via multimodal enter streams.

In contrast to video understanding duties (e.g., video classification and retrieval) the place the important thing problem lies in processing and understanding multimodal enter movies, the duty of multimodal video captioning contains the extra problem of producing grounded captions. Essentially the most broadly adopted strategy for this job is to coach an encoder-decoder community collectively utilizing manually annotated knowledge. Nonetheless, as a consequence of a scarcity of large-scale, manually annotated knowledge, the duty of annotating grounded captions for movies is labor intensive and, in lots of circumstances, impractical. Earlier analysis resembling VideoBERT and CoMVT pre-train their fashions on unlabelled movies by leveraging computerized speech recognition (ASR). Nonetheless, such fashions usually can not generate pure language sentences as a result of they lack a decoder, and thus solely the video encoder is transferred to the downstream duties.

In “Finish-to-Finish Generative Pre-training for Multimodal Video Captioning”, printed at CVPR 2022, we introduce a novel pre-training framework for multimodal video captioning. This framework, which we name multimodal video generative pre-training or MV-GPT, collectively trains a multimodal video encoder and a sentence decoder from unlabelled movies by leveraging a future utterance because the goal textual content and formulating a novel bi-directional technology job. We display that MV-GPT successfully transfers to multimodal video captioning, reaching state-of-the-art outcomes on numerous benchmarks. Moreover, the multimodal video encoder is aggressive for a number of video understanding duties, resembling VideoQA, text-video retrieval, and motion recognition.

Future Utterance as an Further Textual content Sign

Sometimes, every coaching video clip for multimodal video captioning is related to two completely different texts: (1) a speech transcript that’s aligned with the clip as part of the multimodal enter stream, and (2) a goal caption, which is usually manually annotated. The encoder learns to fuse info from the transcript with visible contents, and the goal caption is used to coach the decoder for technology. Nonetheless, within the case of unlabelled movies, every video clip comes solely with a transcript from ASR, and not using a manually annotated goal caption. Furthermore, we can not use the identical textual content (the ASR transcript) for the encoder enter and decoder goal, for the reason that technology of the goal would then be trivial.

MV-GPT circumvents this problem by leveraging a future utterance as an extra textual content sign and enabling joint pre-training of the encoder and decoder. Nonetheless, coaching a mannequin to generate future utterances which might be usually not grounded within the enter content material just isn’t best. So we apply a novel bi-directional technology loss to strengthen the connection to the enter.

Bi-directional Technology Loss

The problem of non-grounded textual content technology is mitigated by formulating a bi-directional technology loss that features ahead and backward technology. Ahead technology produces future utterances given visible frames and their corresponding transcripts and permits the mannequin to study to fuse the visible content material with its corresponding transcript. Backward technology takes the visible frames and future utterances to coach the mannequin to generate a transcript that comprises extra grounded textual content of the video clip. Bi-directional technology loss in MV-GPT permits the encoder and the decoder to be skilled to deal with visually grounded texts.

Bi-directional technology in MV-GPT. A mannequin is skilled with two technology losses. In ahead technology, the mannequin generates a future utterance (blue containers) given the frames and the current utterance (purple containers), whereas the current is generated from the longer term utterance in backward technology. Two particular beginning-of-sentence tokens ([BOS-F] and [BOS-B]) provoke ahead and backward technology for the decoder.

Outcomes on Multimodal Video Captioning

We evaluate MV-GPT to current pre-training losses utilizing the identical mannequin structure, on YouCook2 with normal analysis metrics (Bleu-4, Cider, Meteor and Rouge-L). Whereas all pre-training strategies enhance captioning performances, it’s vital to pre-train the decoder collectively to enhance mannequin efficiency. We display that MV-GPT outperforms the earlier state-of-the-art joint pre-training methodology by over 3.5% with relative positive aspects throughout all 4 metrics.

Pre-training LossPre-trained ElementsBleu-4Cider MeteorRouge-L
No Pre-trainingN/A13.251.0317.5635.48
CoMVTEncoder14.461.2418.4637.17
UniVLEncoder + Decoder19.951.9825.2746.81
MV-GPT (ours)Encoder + Decoder21.262.1426.3648.58
MV-GPT efficiency throughout 4 metrics (Bleu-4, Cider, Meteor and Rouge-L) of various pre-training losses on YouCook2. “Pre-trained elements” signifies which elements of the mannequin are pre-trained — solely the encoder or each the encoder and decoder. We reimplement the loss features of current strategies however use our mannequin and coaching methods for a good comparability.

We switch a mannequin pre-trained by MV-GPT to 4 completely different captioning benchmarks: YouCook2, MSR-VTT, ViTT and ActivityNet-Captions. Our mannequin achieves state-of-the-art efficiency on all 4 benchmarks by important margins. For example on the Meteor metric, MV-GPT exhibits over 12% relative enhancements in all 4 benchmarks.

YouCook2MSR-VTTViTTActivityNet-Captions
Finest Baseline22.3529.9011.0010.90
MV-GPT (ours)27.0938.6626.7512.31
Meteor metric scores of one of the best baseline strategies and MV-GPT on 4 benchmarks.

Outcomes on Non-generative Video Understanding Duties

Though MV-GPT is designed to coach a generative mannequin for multimodal video captioning, we additionally discover that our pre-training approach learns a robust multimodal video encoder that may be utilized to a number of video understanding duties, together with VideoQA, text-video retrieval and motion classification. When in comparison with one of the best comparable baseline fashions, the mannequin transferred from MV-GPT exhibits superior efficiency in 5 video understanding benchmarks on their main metrics — i.e., top-1 accuracy for VideoQA and motion classification benchmarks, and recall at 1 for the retrieval benchmark.

ActivityBenchmarkFinest Comparable BaselineMV-GPT
VideoQAMSRVTT-QA41.541.7
ActivityNet-QA38.939.1
Textual content-Video RetrievalMSR-VTT33.737.3
Motion RecognitionKinetics-40078.980.4
Kinetics-60080.682.4
Comparisons of MV-GPT to greatest comparable baseline fashions on 5 video understanding benchmarks. For every dataset we report the broadly used main metric, i.e., MSRVTT-QA and ActivityNet-QA: Prime-1 reply accuracy; MSR-VTT: Recall at 1; and Kinetics: Prime-1 classification accuracy.

Abstract

We introduce MV-GPT, a brand new generative pre-training framework for multimodal video captioning. Our bi-directional generative goal collectively pre-trains a multimodal encoder and a caption decoder through the use of utterances sampled at completely different instances in unlabelled movies. Our pre-trained mannequin achieves state-of-the-art outcomes on a number of video captioning benchmarks and different video understanding duties, particularly VideoQA, video retrieval and motion classification.

Acknowledgements

This analysis was performed by Paul Hongsuck Website positioning, Arsha Nagrani, Anurag Arnab and Cordelia Schmid.

LEAVE A REPLY

Please enter your comment!
Please enter your name here