This paper offers a comprehensive analysis of recent advancements in video inpainting techniques, a critical subset of computer vision and artificial intelligence. As a process that restores or fills in missing or corrupted portions of video sequences with plausible content, video inpainting has evolved significantly with the advent of deep learning methodologies. Despite the plethora of existing methods and their swift development, the landscape remains complex, posing challenges to both novices and established researchers. Our study deconstructs major techniques, their underpinning theories, and their effective applications. Moreover, we conduct an exhaustive comparative study, centering on two often-overlooked dimensions: visual quality and computational efficiency. We adopt a human-centric approach to assess visual quality, enlisting a panel of annotators to evaluate the output of different video inpainting techniques. This provides a nuanced qualitative understanding that complements traditional quantitative metrics. Concurrently, we delve into the computational aspects, comparing inference times and memory demands across a standardized hardware setup. This analysis underscores the balance between quality and efficiency: a critical consideration for practical applications where resources may be constrained. By integrating human validation and computational resource comparison, this survey not only clarifies the present landscape of video inpainting techniques but also charts a course for future explorations in this vibrant and evolving field.
Despite recent advances in video action recognition achieving strong performance on existing benchmarks, these models often lack robustness when faced with natural distribution shifts between training and test data. We propose two novel evaluation methods to assess model resilience to such distribution disparity. One method uses two different datasets collected from different sources and uses one for training and validation, and the other for testing. More precisely, we created dataset splits of HMDB-51 or UCF-101 for training, and Kinetics-400 for testing, using the subset of the classes that are overlapping in both train and test datasets. The other proposed method extracts the feature mean of each class from the target evaluation dataset's training data (i.e. class prototype) and estimates test video prediction as a cosine similarity score between each sample to the class prototypes of each target class. This procedure does not alter model weights using the target dataset and it does not require aligning overlapping classes of two different datasets, thus is a very efficient method to test the model robustness to distribution shifts without prior knowledge of the target distribution. We address the robustness problem by adversarial augmentation training - generating augmented views of videos that are "hard" for the classification model by applying gradient ascent on the augmentation parameters - as well as "curriculum" scheduling the strength of the video augmentations. We experimentally demonstrate the superior performance of the proposed adversarial augmentation approach over baselines across three state-of-the-art action recognition models - TSM, Video Swin Transformer, and Uniformer. The presented work provides critical insight into model robustness to distribution shifts and presents effective techniques to enhance video action recognition performance in a real-world deployment.
Deep learning models have revolutionized various fields, from image recognition to natural language processing, by achieving unprecedented levels of accuracy. However, their increasing energy consumption has raised concerns about their environmental impact, disadvantaging smaller entities in research and exacerbating global energy consumption. In this paper, we explore the trade-off between model accuracy and electricity consumption, proposing a metric that penalizes large consumption of electricity. We conduct a comprehensive study on the electricity consumption of various deep learning models across different GPUs, presenting a detailed analysis of their accuracy-efficiency trade-offs. By evaluating accuracy per unit of electricity consumed, we demonstrate how smaller, more energy-efficient models can significantly expedite research while mitigating environmental concerns. Our results highlight the potential for a more sustainable approach to deep learning, emphasizing the importance of optimizing models for efficiency. This research also contributes to a more equitable research landscape, where smaller entities can compete effectively with larger counterparts. This advocates for the adoption of efficient deep learning practices to reduce electricity consumption, safeguarding the environment for future generations whilst also helping ensure a fairer competitive landscape.
Video understanding has long suffered from reliance on large labeled datasets, motivating research into zero-shot learning. Recent progress in language modeling presents opportunities to advance zero-shot video analysis, but constructing an effective semantic space relating action classes remains challenging. We address this by introducing a novel dataset, Stories, which contains rich textual descriptions for diverse action classes extracted from WikiHow articles. For each class, we extract multi-sentence narratives detailing the necessary steps, scenes, objects, and verbs that characterize the action. This contextual data enables modeling of nuanced relationships between actions, paving the way for zero-shot transfer. We also propose an approach that harnesses Stories to improve feature generation for training zero-shot classification. Without any target dataset fine-tuning, our method achieves new state-of-the-art on multiple benchmarks, improving top-1 accuracy by up to 6.1%. We believe Stories provides a valuable resource that can catalyze progress in zero-shot action recognition. The textual narratives forge connections between seen and unseen classes, overcoming the bottleneck of labeled data that has long impeded advancements in this exciting domain. The data can be found here: https://github.com/kini5gowda/Stories .
Recent advancements in deep learning and computer vision have led to a surge of interest in generating realistic talking heads. This paper presents a comprehensive survey of state-of-the-art methods for talking head generation. We systematically categorises them into four main approaches: image-driven, audio-driven, video-driven and others (including neural radiance fields (NeRF), and 3D-based methods). We provide an in-depth analysis of each method, highlighting their unique contributions, strengths, and limitations. Furthermore, we thoroughly compare publicly available models, evaluating them on key aspects such as inference time and human-rated quality of the generated outputs. Our aim is to provide a clear and concise overview of the current landscape in talking head generation, elucidating the relationships between different approaches and identifying promising directions for future research. This survey will serve as a valuable reference for researchers and practitioners interested in this rapidly evolving field.
In this paper, we address the challenges posed by the substantial training time and memory consumption associated with video transformers, focusing on the ViViT (Video Vision Transformer) model, in particular the Factorised Encoder version, as our baseline for action recognition tasks. The factorised encoder variant follows the late-fusion approach that is adopted by many state of the art approaches. Despite standing out for its favorable speed/accuracy tradeoffs among the different variants of ViViT, its considerable training time and memory requirements still pose a significant barrier to entry. Our method is designed to lower this barrier and is based on the idea of freezing the spatial transformer during training. This leads to a low accuracy model if naively done. But we show that by (1) appropriately initializing the temporal transformer (a module responsible for processing temporal information) (2) introducing a compact adapter model connecting frozen spatial representations ((a module that selectively focuses on regions of the input image) to the temporal transformer, we can enjoy the benefits of freezing the spatial transformer without sacrificing accuracy. Through extensive experimentation over 6 benchmarks, we demonstrate that our proposed training strategy significantly reduces training costs (by $\sim 50\%$) and memory consumption while maintaining or slightly improving performance by up to 1.79\% compared to the baseline model. Our approach additionally unlocks the capability to utilize larger image transformer models as our spatial transformer and access more frames with the same memory consumption.
Generalized Zero-Shot Learning (GZSL) has emerged as a pivotal research domain in computer vision, owing to its capability to recognize objects that have not been seen during training. Despite the significant progress achieved by generative techniques in converting traditional GZSL to fully supervised learning, they tend to generate a large number of synthetic features that are often redundant, thereby increasing training time and decreasing accuracy. To address this issue, this paper proposes a novel approach for synthetic feature selection using reinforcement learning. In particular, we propose a transformer-based selector that is trained through proximal policy optimization (PPO) to select synthetic features based on the validation classification accuracy of the seen classes, which serves as a reward. The proposed method is model-agnostic and data-agnostic, making it applicable to both images and videos and versatile for diverse applications. Our experimental results demonstrate the superiority of our approach over existing feature-generating methods, yielding improved overall performance on multiple benchmarks.
Understanding the steps required to perform a task is an important skill for AI systems. Learning these steps from instructional videos involves two subproblems: (i) identifying the temporal boundary of sequentially occurring segments and (ii) summarizing these steps in natural language. We refer to this task as Procedure Segmentation and Summarization (PSS). In this paper, we take a closer look at PSS and propose three fundamental improvements over current methods. The segmentation task is critical, as generating a correct summary requires each step of the procedure to be correctly identified. However, current segmentation metrics often overestimate the segmentation quality because they do not consider the temporal order of segments. In our first contribution, we propose a new segmentation metric that takes into account the order of segments, giving a more reliable measure of the accuracy of a given predicted segmentation. Current PSS methods are typically trained by proposing segments, matching them with the ground truth and computing a loss. However, much like segmentation metrics, existing matching algorithms do not consider the temporal order of the mapping between candidate segments and the ground truth. In our second contribution, we propose a matching algorithm that constrains the temporal order of segment mapping, and is also differentiable. Lastly, we introduce multi-modal feature training for PSS, which further improves segmentation. We evaluate our approach on two instructional video datasets (YouCook2 and Tasty) and observe an improvement over the state-of-the-art of $\sim7\%$ and $\sim2.5\%$ for procedure segmentation and summarization, respectively.
We address the problem of data augmentation for video action recognition. Standard augmentation strategies in video are hand-designed and sample the space of possible augmented data points either at random, without knowing which augmented points will be better, or through heuristics. We propose to learn what makes a good video for action recognition and select only high-quality samples for augmentation. In particular, we choose video compositing of a foreground and a background video as the data augmentation process, which results in diverse and realistic new samples. We learn which pairs of videos to augment without having to actually composite them. This reduces the space of possible augmentations, which has two advantages: it saves computational cost and increases the accuracy of the final trained classifier, as the augmented pairs are of higher quality than average. We present experimental results on the entire spectrum of training settings: few-shot, semi-supervised and fully supervised. We observe consistent improvements across all of them over prior work and baselines on Kinetics, UCF101, HMDB51, and achieve a new state-of-the-art on settings with limited data. We see improvements of up to 8.6% in the semi-supervised setting.
We address the problem of capturing temporal information for video classification in 2D networks, without increasing computational cost. Existing approaches focus on modifying the architecture of 2D networks (e.g. by including filters in the temporal dimension to turn them into 3D networks, or using optical flow, etc.), which increases computation cost. Instead, we propose a novel sampling strategy, where we re-order the channels of the input video, to capture short-term frame-to-frame changes. We observe that without bells and whistles, the proposed sampling strategy improves performance on multiple architectures (e.g. TSN, TRN, and TSM) and datasets (CATER, Something-Something-V1 and V2), up to 24% over the baseline of using the standard video input. In addition, our sampling strategies do not require training from scratch and do not increase the computational cost of training and testing. Given the generality of the results and the flexibility of the approach, we hope this can be widely useful to the video understanding community. Code is available at https://github.com/kiyoon/PyVideoAI.