Yue Zhao

Yue Zhao, Ishan Misra, Philipp Krähenbühl et al. · meta-ai

We introduce LaViLa, a new approach to learning video-language representations by leveraging Large Language Models (LLMs). We repurpose pre-trained LLMs to be conditioned on visual input, and finetune them to create automatic video narrators. Our auto-generated narrations offer a number of advantages, including dense coverage of long videos, better temporal synchronization of the visual information and text, and much higher diversity of text. The video-text embedding learned contrastively with these additional auto-generated narrations outperforms the previous state-of-the-art on multiple first-person and third-person video tasks, both in zero-shot and finetuned setups. Most notably, LaViLa obtains an absolute gain of 10.1% on EGTEA classification and 5.9% Epic-Kitchens-100 multi-instance retrieval benchmarks. Furthermore, LaViLa trained with only half the narrations from the Ego4D dataset outperforms baseline models trained on the full set, and shows positive scaling behavior on increasing pre-training data and model size.

Yue Zhao, Philipp Krähenbühl

Videos are big, complex to pre-process, and slow to train on. State-of-the-art large-scale video models are trained on clusters of 32 or more GPUs for several days. As a consequence, academia largely ceded the training of large video models to industry. In this paper, we show how to still train a state-of-the-art video model on a single machine with eight consumer-grade GPUs in a day. We identify three bottlenecks, IO, CPU, and GPU computation, and optimize each. The result is a highly efficient video training pipeline. For comparable architectures, our pipeline achieves higher accuracies with $\frac{1}{8}$ of the computation compared to prior work. Code is available at https://github.com/zhaoyue-zephyrus/AVION.

Yue Zhao, Philipp Krähenbühl

Streaming video recognition reasons about objects and their actions in every frame of a video. A good streaming recognition model captures both long-term dynamics and short-term changes of video. Unfortunately, in most existing methods, the computational complexity grows linearly or quadratically with the length of the considered dynamics. This issue is particularly pronounced in transformer-based architectures. To address this issue, we reformulate the cross-attention in a video transformer through the lens of kernel and apply two kinds of temporal smoothing kernel: A box kernel or a Laplace kernel. The resulting streaming attention reuses much of the computation from frame to frame, and only requires a constant time update each frame. Based on this idea, we build TeSTra, a Temporal Smoothing Transformer, that takes in arbitrarily long inputs with constant caching and computing overhead. Specifically, it runs $6\times$ faster than equivalent sliding-window based transformers with 2,048 frames in a streaming setting. Furthermore, thanks to the increased temporal span, TeSTra achieves state-of-the-art results on THUMOS'14 and EPIC-Kitchen-100, two standard online action detection and action anticipation datasets. A real-time version of TeSTra outperforms all but one prior approaches on the THUMOS'14 dataset.

Yufeng Shi, Shuhuang Chen, Xinge You et al.

Mapping X-ray images, radiology reports, and other medical data as binary codes in the common space, which can assist clinicians to retrieve pathology-related data from heterogeneous modalities (i.e., hashing-based cross-modal medical data retrieval), provides a new view to promot computeraided diagnosis. Nevertheless, there remains a barrier to boost medical retrieval accuracy: how to reveal the ambiguous semantics of medical data without the distraction of superfluous information. To circumvent this drawback, we propose Deep Supervised Information Bottleneck Hashing (DSIBH), which effectively strengthens the discriminability of hash codes. Specifically, the Deep Deterministic Information Bottleneck (Yu, Yu, and Principe 2021) for single modality is extended to the cross-modal scenario. Benefiting from this, the superfluous information is reduced, which facilitates the discriminability of hash codes. Experimental results demonstrate the superior accuracy of the proposed DSIBH compared with state-of-the-arts in cross-modal medical data retrieval tasks.

Long Zhao, Nitesh B. Gundavarapu, Liangzhe Yuan et al. · deepmind

We introduce VideoPrism, a general-purpose video encoder that tackles diverse video understanding tasks with a single frozen model. We pretrain VideoPrism on a heterogeneous corpus containing 36M high-quality video-caption pairs and 582M video clips with noisy parallel text (e.g., ASR transcripts). The pretraining approach improves upon masked autoencoding by global-local distillation of semantic video embeddings and a token shuffling scheme, enabling VideoPrism to focus primarily on the video modality while leveraging the invaluable text associated with videos. We extensively test VideoPrism on four broad groups of video understanding tasks, from web video question answering to CV for science, achieving state-of-the-art performance on 31 out of 33 video understanding benchmarks. Our models are released at https://github.com/google-deepmind/videoprism.

Adam Polyak, Amit Zohar, Andrew Brown et al. · meta-ai

We present Movie Gen, a cast of foundation models that generates high-quality, 1080p HD videos with different aspect ratios and synchronized audio. We also show additional capabilities such as precise instruction-based video editing and generation of personalized videos based on a user's image. Our models set a new state-of-the-art on multiple tasks: text-to-video synthesis, video personalization, video editing, video-to-audio generation, and text-to-audio generation. Our largest video generation model is a 30B parameter transformer trained with a maximum context length of 73K video tokens, corresponding to a generated video of 16 seconds at 16 frames-per-second. We show multiple technical innovations and simplifications on the architecture, latent spaces, training objectives and recipes, data curation, evaluation protocols, parallelization techniques, and inference optimizations that allow us to reap the benefits of scaling pre-training data, model size, and training compute for training large scale media generation models. We hope this paper helps the research community to accelerate progress and innovation in media generation models. All videos from this paper are available at https://go.fb.me/MovieGenResearchVideos.

Kecheng Chen, Jiayu Sun, Jiang Shen et al.

Being low-level radiation exposure and less harmful to health, low-dose computed tomography (LDCT) has been widely adopted in the early screening of lung cancer and COVID-19. LDCT images inevitably suffer from the degradation problem caused by complex noises. It was reported that deep learning (DL)-based LDCT denoising methods using convolutional neural network (CNN) achieved impressive denoising performance. Although most existing DL-based methods (e.g., encoder-decoder framework) can implicitly utilize non-local and contextual information via downsampling operator and 3D CNN, the explicit multi-information (i.e., local, non-local, and contextual) integration may not be explored enough. To address this issue, we propose a novel graph convolutional network-based LDCT denoising model, namely GCN-MIF, to explicitly perform multi-information fusion for denoising purpose. Concretely, by constructing intra- and inter-slice graph, the graph convolutional network is introduced to leverage the non-local and contextual relationships among pixels. The traditional CNN is adopted for the extraction of local information. Finally, the proposed GCN-MIF model fuses all the extracted local, non-local, and contextual information. Extensive experiments show the effectiveness of our proposed GCN-MIF model by quantitative and visualized results. Furthermore, a double-blind reader study on a public clinical dataset is also performed to validate the usability of denoising results in terms of the structural fidelity, the noise suppression, and the overall score. Models and code are available at https://github.com/tonyckc/GCN-MIF_demo.

Karan Dalal, Daniel Koceja, Gashon Hussein et al.

Transformers today still struggle to generate one-minute videos because self-attention layers are inefficient for long context. Alternatives such as Mamba layers struggle with complex multi-scene stories because their hidden states are less expressive. We experiment with Test-Time Training (TTT) layers, whose hidden states themselves can be neural networks, therefore more expressive. Adding TTT layers into a pre-trained Transformer enables it to generate one-minute videos from text storyboards. For proof of concept, we curate a dataset based on Tom and Jerry cartoons. Compared to baselines such as Mamba~2, Gated DeltaNet, and sliding-window attention layers, TTT layers generate much more coherent videos that tell complex stories, leading by 34 Elo points in a human evaluation of 100 videos per method. Although promising, results still contain artifacts, likely due to the limited capability of the pre-trained 5B model. The efficiency of our implementation can also be improved. We have only experimented with one-minute videos due to resource constraints, but the approach can be extended to longer videos and more complex stories. Sample videos, code and annotations are available at: https://test-time-training.github.io/video-dit

Yue Zhao, Long Zhao, Xingyi Zhou et al. · deepmind

The recent advance in vision-language models is largely attributed to the abundance of image-text data. We aim to replicate this success for video-language models, but there simply is not enough human-curated video-text data available. We thus resort to fine-tuning a video-language model from a strong image-language baseline with synthesized instructional data. The resulting video model by video-instruction-tuning (VIIT) is then used to auto-label millions of videos to generate high-quality captions. We show the adapted video-language model performs well on a wide range of video-language benchmarks. For instance, it surpasses the best prior result on open-ended NExT-QA by 2.8%. Besides, our model generates detailed descriptions for previously unseen videos, which provide better textual supervision than existing methods. Experiments show that a video-language dual-encoder model contrastively trained on these auto-generated captions is 3.8% better than the strongest baseline that also leverages vision-language models. Our best model outperforms state-of-the-art methods on MSR-VTT zero-shot text-to-video retrieval by 6%. As a side product, we generate the largest video caption dataset to date.