Generating temporally-consistent high-fidelity videos can be computationally expensive, especially over longer temporal spans. More-recent Diffusion Transformers (DiTs) -- despite making significant headway in this context -- have only heightened such challenges as they rely on larger models and heavier attention mechanisms, resulting in slower inference speeds. In this paper, we introduce a training-free method to accelerate video DiTs, termed Adaptive Caching (AdaCache), which is motivated by the fact that "not all videos are created equal": meaning, some videos require fewer denoising steps to attain a reasonable quality than others. Building on this, we not only cache computations through the diffusion process, but also devise a caching schedule tailored to each video generation, maximizing the quality-latency trade-off. We further introduce a Motion Regularization (MoReg) scheme to utilize video information within AdaCache, essentially controlling the compute allocation based on motion content. Altogether, our plug-and-play contributions grant significant inference speedups (e.g. up to 4.7x on Open-Sora 720p - 2s video generation) without sacrificing the generation quality, across multiple video DiT baselines.

Vision Language Models (VLMs) have recently been leveraged to generate robotic actions, forming Vision-Language-Action (VLA) models. However, directly adapting a pretrained VLM for robotic control remains challenging, particularly when constrained by a limited number of robot demonstrations. In this work, we introduce LLaRA: Large Language and Robotics Assistant, a framework that formulates robot action policy as visuo-textual conversations and enables an efficient transfer of a pretrained VLM into a powerful VLA, motivated by the success of visual instruction tuning in Computer Vision. First, we present an automated pipeline to generate conversation-style instruction tuning data for robots from existing behavior cloning datasets, aligning robotic actions with image pixel coordinates. Further, we enhance this dataset in a self-supervised manner by defining six auxiliary tasks, without requiring any additional action annotations. We show that a VLM finetuned with a limited amount of such datasets can produce meaningful action decisions for robotic control. Through experiments across multiple simulated and real-world tasks, we demonstrate that LLaRA achieves state-of-the-art performance while preserving the generalization capabilities of large language models. The code, datasets, and pretrained models are available at https://github.com/LostXine/LLaRA.

Large Language Models (LLMs) have allowed recent LLM-based approaches to achieve excellent performance on long-video understanding benchmarks. We investigate how extensive world knowledge and strong reasoning skills of underlying LLMs influence this strong performance. Surprisingly, we discover that LLM-based approaches can yield surprisingly good accuracy on long-video tasks with limited video information, sometimes even with no video-specific information. Building on this, we explore injecting video-specific information into an LLM-based framework. We utilize off-the-shelf vision tools to extract three object-centric information modalities from videos, and then leverage natural language as a medium for fusing this information. Our resulting Multimodal Video Understanding (MVU) framework demonstrates state-of-the-art performance across multiple video understanding benchmarks. Strong performance also on robotics domain tasks establishes its strong generality. Code: github.com/kahnchana/mvu

Vision-Language models (VLMs) have excelled in the image-domain- especially in zero-shot settings- thanks to the availability of vast pretraining data (i.e., paired image-text samples). However for videos, such paired data is not as abundant. Therefore, video- VLMs are usually designed by adapting pretrained image- VLMs to the video-domain, instead of training from scratch. All such recipes rely on aug-menting visual embeddings with temporal information (i.e., image -+ video), often keeping text embeddings unchanged or even being discarded. In this paper, we argue the contrary, that better video- VLMs can be designed by focusing more on augmenting text, rather than visual information. More specifically, we introduce Video-conditioned Text Representations (Vi c TR): a form of text embeddings optimized w.r.t. vi-sual embeddings, creating a more-flexible contrastive latent space. Our model canfurther make use offreely-available semantic information, in the form of visually- grounded aux-iliary text (e.g. object or scene information). We evaluate our model on few-shot, zero-shot (HMDB-51, UCF-10l), short-form (Kinetics-400) and long-form (Charades) activ-ity recognition benchmarks, showing strong performance among video-VLMs.

Vision Transformers (ViTs) have recently become the state-of-the-art across many computer vision tasks. In contrast to convolutional networks (CNNs), ViTs enable global information sharing even within shallow layers of a network, i.e., among high-resolution features. However, this perk was later overlooked with the success of pyramid architectures such as Swin Transformer, which show better performance-complexity trade-offs. In this paper, we present a simple and efficient add-on component (termed GrafT) that considers global dependencies and multi-scale information throughout the network, in both high- and low-resolution features alike. It has the flexibility of branching out at arbitrary depths and shares most of the parameters and computations of the backbone. GrafT shows consistent gains over various well-known models which includes both hybrid and pure Transformer types, both homogeneous and pyramid structures, and various self-attention methods. In particular, it largely benefits mobile-size models by providing high-level semantics. On the ImageNet-1k dataset, GrafT delivers +3.9%, +1.4%, and +1.9% top-1 accuracy improvement to DeiT-T, Swin-T, and MobViTXXS, respectively. Our code and models are at https://github.com/jongwoopark7978/Grafting-Vision-Transformer.

Temporal Activity Detection aims to predict activity classes per frame, in contrast to video-level predictions in Activity Classification (i.e., Activity Recognition). Due to the expensive frame-level annotations required for detection, the scale of detection datasets is limited. Thus, commonly, previous work on temporal activity detection resorts to fine-tuning a classification model pretrained on large-scale classification datasets (e.g., Kinetics-400). However, such pretrained models are not ideal for downstream detection, due to the disparity between the pretraining and the downstream fine-tuning tasks. In this work, we propose a novel weakly-guided self-supervised pretraining method for detection. We leverage weak labels (classification) to introduce a self-supervised pretext task (detection) by generating frame-level pseudo labels, multi-action frames, and action segments. Simply put, we design a detection task similar to downstream, on large-scale classification data, without extra annotations. We show that the models pretrained with the proposed weakly-guided self-supervised detection task outperform prior work on multiple challenging activity detection benchmarks, including Charades and MultiTHUMOS. Our extensive ablations further provide insights on when and how to use the proposed models for activity detection. Code is available at github.com/kkahatapitiya/SSDet.

We propose Token Turing Machines (TTM), a sequential, autoregressive Transformer model with memory for real-world sequential visual understanding. Our model is inspired by the seminal Neural Turing Machine, and has an external memory consisting of a set of tokens which summarise the previous history (i.e., frames). This memory is efficiently addressed, read and written using a Transformer as the processing unit/controller at each step. The model's memory module ensures that a new observation will only be processed with the contents of the memory (and not the entire history), meaning that it can efficiently process long sequences with a bounded computational cost at each step. We show that TTM outperforms other alternatives, such as other Transformer models designed for long sequences and recurrent neural networks, on two real-world sequential visual understanding tasks: online temporal activity detection from videos and vision-based robot action policy learning. Code is publicly available at: https://github.com/google-research/scenic/tree/main/scenic/projects/token.turing.

Modeling visual data as tokens (i.e., image patches) using attention mechanisms, feed-forward networks or convolutions has been highly effective in recent years. Such methods usually have a common pipeline: a tokenization method, followed by a set of layers/blocks for information mixing, both within and among tokens. When image patches are converted into tokens, they are often flattened, discarding the spatial structure within each patch. As a result, any processing that follows (eg: multi-head self-attention) may fail to recover and/or benefit from such information. In this paper, we argue that models can have significant gains when spatial structure is preserved during tokenization, and is explicitly used during the mixing stage. We propose two key contributions: (1) Structure-aware Tokenization and, (2) Structure-aware Mixing, both of which can be combined with existing models with minimal effort. We introduce a family of models (SWAT), showing improvements over the likes of DeiT, MLP-Mixer and Swin Transformer, across multiple benchmarks including ImageNet classification and ADE20K segmentation. Our code is available at github.com/kkahatapitiya/SWAT.

Reinforcement Learning (RL) can be considered as a sequence modeling task, where an agent employs a sequence of past state-action-reward experiences to predict a sequence of future actions. In this work, we propose State-Action-Reward Transformer (StARformer), a Transformer architecture for robot learning with image inputs, which explicitly models short-term state-action-reward representations (StAR-representations), essentially introducing a Markovian-like inductive bias to improve long-term modeling. StARformer first extracts StAR-representations using self-attending patches of image states, action, and reward tokens within a short temporal window. These StAR-representations are combined with pure image state representations, extracted as convolutional features, to perform self-attention over the whole sequence. Our experimental results show that StARformer outperforms the state-of-the-art Transformer-based method on image-based Atari and DeepMind Control Suite benchmarks, under both offline-RL and imitation learning settings. We find that models can benefit from our combination of patch-wise and convolutional image embeddings. StARformer is also more compliant with longer sequences of inputs than the baseline method. Finally, we demonstrate how StARformer can be successfully applied to a real-world robot imitation learning setting via a human-following task.

Action detection is a significant and challenging task, especially in densely-labelled datasets of untrimmed videos. Such data consist of complex temporal relations including composite or co-occurring actions. To detect actions in these complex settings, it is critical to capture both shortterm and long-term temporal information efficiently. To this end, we propose a novel ‘ConvTransformer’ network for action detection: MS-TCT11Code/Models: https://github.com/dairui01/MS-TCT. This network comprises of three main components: (1) a Temporal Encoder module which explores global and local temporal relations at multiple temporal resolutions, (2) a Temporal Scale Mixer module which effectively fuses multi-scale features, creating a unified feature representation, and (3) a Classification module which learns a center-relative position of each action instance in time, and predicts frame-level classification scores. Our experimental results on multiple challenging datasets such as Charades, TSU and MultiTHUMOS, validate the effectiveness of the proposed method, which outperforms the state-of-the-art methods on all three datasets.

In this paper, we introduce Coarse-Fine Networks, a twostream architecture which benefits from different abstractions of temporal resolution to learn better video representations for long-term motion. Traditional Video models process inputs at one (or few) fixed temporal resolution without any dynamic frame selection. However, we argue that, processing multiple temporal resolutions of the input and doing so dynamically by learning to estimate the importance of each frame can largely improve video representations, specially in the domain of temporal activity localization. To this end, we propose (1) ‘Grid Pool’, a learned temporal downsampling layer to extract coarse features, and, (2) ‘Multi-stage Fusion’, a spatio-temporal attention mechanism to fuse a finegrained context with the coarse features. We show that our method outperforms the state-of-the-arts for action detection in public datasets including Charades with a significantly reduced compute and memory footprint. The code is available at https://github.com/kkahatapitiya/Coarse-Fine-Networks.

Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance in many computer vision tasks over the years. However, this comes at the cost of heavy computation and memory intensive network designs, suggesting potential improvements in efficiency. Convolutional layers of CNNs partly account for such an inefficiency, as they are known to learn redundant features. In this work, we exploit this redundancy, observing it as the correlation between convolutional filters of a layer, and propose an alternative approach to reproduce it efficiently. The proposed `LinearConv' layer learns a set of orthogonal filters, and a set of coefficients that linearly combines them to introduce a controlled redundancy. We introduce a correlation-based regularization loss to achieve such flexibility over redundancy, and control the number of parameters in turn. This is designed as a plug-and-play layer to conveniently replace a conventional convolutional layer, without any additional changes required in the network architecture or the hyper-parameter settings. Our experiments verify that LinearConv models achieve a performance on-par with their counter-parts, with almost a 50% reduction in parameters on average, and the same computational requirement and speed at inference. Source is available at https://github.com/kkahatapitiya/LinearConv.

Learning a particular task from a dataset, samples in which originate from diverse contexts, is challenging, and usually addressed by deepening or widening standard neural networks. As opposed to conventional network widening, multipath architectures restrict the quadratic increment of complexity to a linear scale. However, existing multi-column/path networks or model ensembling methods do not consider any feature-dependent allocation of parallel resources, and therefore, tend to learn redundant features. Given a layer in a multi-path network, if we restrict each path to learn a context-specific set of features and introduce a mechanism to intelligently allocate incoming feature maps to such paths, each path can specialize in a certain context, reducing the redundancy and improving the quality of extracted features. This eventually leads to better-optimized usage of parallel resources. To do this, we propose inserting feature-dependent cross-connections between parallel sets of feature maps in successive layers. The weighting coefficients of these cross-connections are computed from the input features of the particular layer. Our multi-path networks show improved image recognition accuracy at a similar complexity compared to conventional and state-of-the-art methods for deepening, widening and adaptive feature extracting, in both small and large scale datasets.

This paper addresses a classical problem in random matrix theory-finding the distribution of the maximum eigen-value of the correlated Wishart unitary ensemble. In particular, we derive a new exact expression for the cumulative distribution function (c.d. f.) of the maximum eigen-value of a 2 × 2 correlated non-central Wishart matrix with rank-l mean. By using this new result, we derive the exact outage probability of 2 × 2 multiple-input multiple-output maximum-ratio-combining (MIMO-MRC) in Rician fading with transmit correlation and a strong line-of-sight (LoS) component (rank-l channel mean). We also show that the outage performance is affected by the relative alignment of the eigen-spaces of the mean and correlation matrices. In general, when the LoS path aligns with the least eigenvector of the correlation matrix, in the high transmit signal-to-noise ratio (SNR) regime, the outage gradually improves with the increasing correlation. Moreover, we show that as K (Rician factor) grows large, the outage event can be approximately characterized by the c.d.f. of a certain Gaussian random variable.

Occlusion removal is an interesting application of image enhancement, for which, existing work suggests manually-annotated or domain-specific occlusion removal. No work tries to address automatic occlusion detection and removal as a context-aware generic problem. In this paper, we present a novel methodology to identify objects that do not relate to the image context as occlusions and remove them, reconstructing the space occupied coherently. The proposed system detects occlusions by considering the relation between foreground and background object classes represented as vector embeddings, and removes them through inpainting. We test our system on COCO-Stuff dataset and conduct a user study to establish a baseline in context-aware automatic occlusion removal.

Fitness related wearables have become ubiquitous in the recent past. Nevertheless, short battery life of these devices is still a pressing issue. Limited battery capacity in small form factor and power hungry continuous monitoring of accelerometer have been significant concerns in this regard. To address these issues we propose a novel low-power step counting solution based on an Electromagnetic energy harvesting mechanism. Extremely simple nature of the step counter removes the requirement of any step detection algorithm, thereby reducing the power consumption, while the energy harvester generates a portion of energy requirement prolonging the battery life.