Sparse video tubes for joint video and picture imaginative and prescient transformers – Google AI Weblog

Video understanding is a difficult downside that requires reasoning about each spatial data (e.g., for objects in a scene, together with their places and relations) and temporal information for actions or occasions proven in a video. There are a lot of video understanding purposes and duties, similar to understanding the semantic content of web videos and robotic notion. Nevertheless, present works, similar to ViViT and TimeSFormer, densely course of the video and require vital compute, particularly as mannequin dimension plus video size and backbone improve.

In “Rethinking Video ViTs: Sparse Video Tubes for Joint Image and Video Learning”, to be introduced at CVPR 2023, we introduce a easy method that turns a Vision Transformer (ViT) mannequin picture encoder into an environment friendly video spine utilizing sparse video tubes (learnable visible representations of samples from the video) to scale back the mannequin’s compute wants. This strategy can seamlessly course of each photographs and movies, which permits it to leverage each picture and video knowledge sources throughout coaching. This coaching additional permits our sparse tubes ViT mannequin to coalesce picture and video backbones collectively to serve a twin function as both a picture or video spine (or each), relying on the enter. We display that this mannequin is scalable, may be tailored to massive pre-trained ViTs with out requiring full fine-tuning, and achieves state-of-the-art outcomes throughout many video classification benchmarks.

Utilizing sparse video tubes to pattern a video, mixed with a typical ViT encoder, results in an environment friendly visible illustration that may be seamlessly shared with picture inputs.

Constructing a joint image-video spine

Our sparse tube ViT makes use of a typical ViT spine, consisting of a stack of Transformer layers, that processes video data. Earlier strategies, similar to ViViT, densely tokenize the video after which apply factorized attention, i.e., the eye weights for every token are computed individually for the temporal and spatial dimensions. In the usual ViT structure, self-attention is computed over the entire token sequence. When utilizing movies as enter, token sequences turn out to be fairly lengthy, which might make this computation sluggish. As an alternative, within the methodology we suggest, the video is sparsely sampled utilizing video tubes, that are 3D learnable visible representations of varied styles and sizes (described in additional element beneath) from the video. These tubes are used to sparsely pattern the video utilizing a large temporal stride, i.e., when a tube kernel is just utilized to a couple places within the video, reasonably than each pixel.

By sparsely sampling the video tubes, we are able to use the identical international self-attention module, reasonably than factorized consideration like ViViT. We experimentally present that the addition of factorized consideration layers can hurt the efficiency because of the uninitialized weights. This single stack of transformer layers within the ViT spine additionally permits higher sharing of the weights and improves efficiency. Sparse video tube sampling is finished by utilizing a big spatial and temporal stride that selects tokens on a hard and fast grid. The massive stride reduces the variety of tokens within the full community, whereas nonetheless capturing each spatial and temporal data and enabling the environment friendly processing of all tokens.

Sparse video tubes

Video tubes are 3D grid-based cuboids that may have totally different shapes or classes and seize totally different data with strides and beginning places that may overlap. Within the mannequin, we use three distinct tube shapes that seize: (1) solely spatial data (leading to a set of 2D picture patches), (2) lengthy temporal data (over a small spatial space), and (3) each spatial and temporal data equally. Tubes that seize solely spatial data may be utilized to each picture and video inputs. Tubes that seize lengthy temporal data or each temporal and spatial data equally are solely utilized to video inputs. Relying on the enter video dimension, the three tube shapes are utilized to the mannequin a number of occasions to generate tokens.

A hard and fast place embedding, which captures the worldwide location of every tube (together with any strides, offsets, and many others.) relative to all the opposite tubes, is utilized to the video tubes. Totally different from the earlier realized place embeddings, this mounted one higher permits sparse, overlapping sampling. Capturing the worldwide location of the tube helps the mannequin know the place every got here from, which is very useful when tubes overlap or are sampled from distant video places. Subsequent, the tube options are concatenated collectively to kind a set of N tokens. These tokens are processed by a typical ViT encoder. Lastly, we apply an consideration pooling to compress all of the tokens right into a single illustration and enter to a totally linked (FC) layer to make the classification (e.g., taking part in soccer, swimming, and many others.).

Our video ViT mannequin works by sampling sparse video tubes from the video (proven on the backside) to allow both or each picture or video inputs to be seamlessly processed. These tubes have totally different shapes and seize totally different video options. Tube 1 (yellow) solely captures spatial data, leading to a set of 2D patches that may be utilized to picture inputs. Tube 2 (pink) captures temporal data and a few spatial data and tube 3 (inexperienced) equally captures each temporal and spatial data (i.e., the spatial dimension of the tube x and y are the identical because the variety of frames t). Tubes 2 and three can solely be utilized to video inputs. The place embedding is added to all of the tube options.

Scaling video ViTs

The method of constructing video backbones is computationally intensive, however our sparse tube ViT mannequin permits computationally environment friendly scaling of video fashions, leveraging beforehand educated picture backbones. Since picture backbones may be tailored to a video spine, massive picture backbones may be was massive video backbones. Extra particularly, one can switch the realized video characteristic representations from a small tube ViT to a big pre-trained picture ViT and practice the ensuing mannequin with video knowledge for only some steps, versus a full coaching from scratch.

Our strategy permits scaling a sparse tube ViT in a extra environment friendly method. Particularly, the video options from a small video ViT (high community) may be transferred to a big, pre-trained picture ViT (backside community), and additional fine-tuned. This requires fewer coaching steps to realize robust efficiency with the big mannequin. That is useful as massive video fashions could be prohibitively costly to coach from scratch.

Outcomes

We consider our sparse tube ViT strategy utilizing Kinetics-400 (proven beneath), Kinetics-600 and Kinetics-700 datasets and examine its efficiency to an extended listing of prior strategies. We discover that our strategy outperforms all prior strategies. Importantly, it outperforms all state-of-the-art strategies educated collectively on picture+video datasets.

Efficiency in comparison with a number of prior works on the favored Kinetics-400 video dataset. Our sparse tube ViT outperforms state-of-the-art strategies.

Moreover, we check our sparse tube ViT mannequin on the Something-Something V2 dataset, which is usually used to guage extra dynamic actions, and likewise report that it outperforms all prior state-of-the-art approaches.

Efficiency on the One thing-One thing V2 video dataset.

Visualizing some realized kernels

It’s attention-grabbing to grasp what sort of rudimentary options are being realized by the proposed mannequin. We visualize them beneath, displaying each the 2D patches, that are shared for each photographs and movies, and video tubes. These visualizations present the 2D or 3D data being captured by the projection layer. For instance, within the 2D patches, varied widespread options, like edges and colours, are detected, whereas the 3D tubes seize fundamental shapes and the way they might change over time.

Visualizations of patches and tubes realized the sparse tube ViT mannequin. High row are the 2D patches and the remaining two rows are snapshots from the realized video tubes. The tubes present every patch for the 8 or 4 frames to which they’re utilized.

Conclusions

Now we have introduced a brand new sparse tube ViT, which might flip a ViT encoder into an environment friendly video mannequin, and might seamlessly work with each picture and video inputs. We additionally confirmed that enormous video encoders may be bootstrapped from small video encoders and image-only ViTs. Our strategy outperforms prior strategies throughout a number of in style video understanding benchmarks. We consider that this easy illustration can facilitate way more environment friendly studying with enter movies, seamlessly incorporate both picture or video inputs and successfully eradicate the bifurcation of picture and video fashions for future multimodal understanding.

Acknowledgements

This work is carried out by AJ Piergiovanni, Weicheng Kuo and Anelia Angelova, who at the moment are at Google DeepMind. We thank Abhijit Ogale, Luowei Zhou, Claire Cui and our colleagues in Google Analysis for his or her useful discussions, feedback, and help.