In this paper, we propose and conduct a comprehensive benchmark on moment localization task, which aims to retrieve a segment that corresponds to a text query from a single untrimmed video. Our study starts from an observation that most moment localization papers report experimental results only on a few datasets in spite of availability of far more benchmarks. Thus, we conduct an extensive benchmark study to measure the performance of representative methods on widely used 7 datasets. Looking further into the details, we pose additional research questions and empirically verify them, including if they rely on unintended biases introduced by specific training data, if advanced visual features trained on classification task transfer well to this task, and if computational cost of each model pays off. With a series of these experiments, we provide multi-faceted evaluation of state-of-the-art moment localization models. Codes are available at \url{https://github.com/snuviplab/MoLEF}.

In Multimodal Large Language Models (MLLMs), a visual projector plays a crucial role in bridging pre-trained vision encoders with LLMs, enabling profound visual understanding while harnessing the LLMs' robust capabilities. Despite the importance of the visual projector, it has been relatively less explored. In this study, we first identify two essential projector properties: (i) flexibility in managing the number of visual tokens, crucial for MLLMs' over-all efficiency, and (ii) preservation of local context from visual features, vital for spatial understanding. Based on these findings, we propose a novel projector design that is both flexible and locality-enhanced, effectively satisfying the two desirable properties. Additionally, we present comprehensive strategies to effectively utilize multiple and multifaceted instruction datasets. Through extensive experiments, we examine the impact of individual design choices. Finally, our proposed MLLM, Honeybee, remarkably outperforms previous state-of-the-art methods across various benchmarks, including MME, MMBench, SEED-Bench, and LLaVA-Bench, achieving significantly higher efficiency. Code and models are available at https://github.com/kakaobrain/honeybee.

We tackle open-world semantic segmentation, which aims at learning to segment arbitrary visual concepts in images, by using only image-text pairs without dense annotations. Existing open-world segmentation methods have shown impressive advances by employing contrastive learning (CL) to learn diverse visual concepts and transferring the learned image-level understanding to the segmentation task. However, these CL-based methods suffer from a train-test discrepancy, since it only considers image-text alignment during training, whereas segmentation requires region-text alignment during testing. In this paper, we proposed a novel Text-grounded Contrastive Learning (TCL) framework that enables a model to directly learn region-text alignment. Our method generates a segmentation mask for a given text, extracts text-grounded image embedding from the masked region, and aligns it with text embedding via TCL. By learning region-text alignment directly, our framework encourages a model to directly improve the quality of generated segmentation masks. In addition, for a rigorous and fair comparison, we present a unified evaluation protocol with widely used 8 semantic segmentation datasets. TCL achieves state-of-the-art zero-shot segmentation performances with large margins in all datasets. Code is available at https://github.com/kakaobrain/tcl.

Image captioning is one of the straightforward tasks that can take advantage of large-scale web-crawled data which provides rich knowledge about the visual world for a captioning model. However, since web-crawled data contains image-text pairs that are aligned at different levels, the inherent noises (e.g., misaligned pairs) make it difficult to learn a precise captioning model. While the filtering strategy can effectively remove noisy data, it leads to a decrease in learnable knowledge and sometimes brings about a new problem of data deficiency. To take the best of both worlds, we propose a Noise-aware Captioning (NoC) framework, which learns rich knowledge from the whole web-crawled data while being less affected by the noises. This is achieved by the proposed alignment-level-controllable captioner, which is learned using alignment levels of the image-text pairs as a control signal during training. The alignment-level-conditioned training allows the model to generate high-quality captions by simply setting the control signal to the desired alignment level at inference time. An in-depth analysis shows the effectiveness of our framework in handling noise. With two tasks of zero-shot captioning and text-to-image retrieval using generated captions (i.e., self-retrieval), we also demonstrate our model can produce high-quality captions in terms of descriptiveness and distinctiveness. The code is available at https://github.com/kakaobrain/noc.

One of the challenges for analyzing video contents (e.g., actions) is high computational cost, especially for the tasks that require processing densely sampled frames in a long video. We present a novel efficient action recognition algorithm, which allocates computational resources adaptively to individual frames depending on their relevance and significance. Specifically, our algorithm adopts LSTM-based policy modules and sequentially estimates the usefulness of each frame based on their intermediate representations. If a certain frame is unlikely to be helpful for recognizing actions, our model stops forwarding the features to the rest of the layers and starts to consider the next sampled frame. We further reduce the computational cost of our approach by introducing a simple yet effective early termination strategy during the inference procedure. We evaluate the proposed algorithm on three public benchmarks: ActivityNet-v1.3, Mini-Kinetics, and THUMOS’14. Our experiments show that the proposed approach achieves outstanding trade-off between accuracy and efficiency in action recognition.

Human-Object Interaction (HOI) detection is the task of identifying a set of (human, object, interaction) triplets from an image. Recent work proposed transformer encoder-decoder architectures that successfully eliminated the need for many hand-designed components in HOI detection through end-to-end training. However, they are limited to single-scale feature resolution, providing suboptimal performance in scenes containing humans, objects, and their interactions with vastly different scales and distances. To tackle this problem, we propose a Multi-Scale TRansformer (MSTR) for HOI detection powered by two novel HOI-aware deformable attention modules called Dual-Entity attention and Entity-conditioned Context attention. While existing deformable attention comes at a huge cost in HOI detection performance, our proposed attention modules of MSTR learn to effectively attend to sampling points that are essential to identify interactions. In experiments, we achieve the new state-of-the-art performance on two HOI detection benchmarks.

We present a novel framework of knowledge distillation that is capable of learning powerful and efficient student models from ensemble teacher networks. Our approach addresses the inherent model capacity issue between teacher and student and aims to maximize benefit from teacher models during distillation by reducing their capacity gap. Specifically, we employ a neural architecture search technique to augment useful structures and operations, where the searched network is appropriate for knowledge distillation towards student models and free from sacrificing its performance by fixing the network capacity. We also introduce an oracle knowledge distillation loss to facilitate model search and distillation using an ensemble-based teacher model, where a student network is learned to imitate oracle performance of the teacher. We perform extensive experiments on the image classification datasets—CIFAR-100 and TinyImageNet—using various networks. We also show that searching for a new student model is effective in both accuracy and memory size and that the searched models often outperform their teacher models thanks to neural architecture search with oracle knowledge distillation.

This paper addresses the problem of text-to-video temporal grounding, which aims to identify the time interval in a video semantically relevant to a text query. We tackle this problem using a novel regression-based model that learns to extract a collection of mid-level features for semantic phrases in a text query, which corresponds to important semantic entities described in the query (e.g., actors, objects, and actions), and reflect bi-modal interactions between the linguistic features of the query and the visual features of the video in multiple levels. The proposed method effectively predicts the target time interval by exploiting contextual information from local to global during bi-modal interactions. Through in-depth ablation studies, we find out that incorporating both local and global context in video and text interactions is crucial to the accurate grounding. Our experiment shows that the proposed method outperforms the state of the arts on Charades-STA and ActivityNet Captions datasets by large margins, 7.44% and 4.61% points at Recall@tIoU=0.5 metric, respectively.

Dense video captioning is an extremely challenging task since accurate and coherent description of events in a video requires holistic understanding of video contents as well as contextual reasoning of individual events. Most existing approaches handle this problem by first detecting event proposals from a video and then captioning on a subset of the proposals. As a result, the generated sentences are prone to be redundant or inconsistent since they fail to consider temporal dependency between events. To tackle this challenge, we propose a novel dense video captioning framework, which models temporal dependency across events in a video explicitly and leverages visual and linguistic context from prior events for coherent storytelling. This objective is achieved by 1) integrating an event sequence generation network to select a sequence of event proposals adaptively, and 2) feeding the sequence of event proposals to our sequential video captioning network, which is trained by reinforcement learning with two-level rewards - at both event and episode levels - for better context modeling. The proposed technique achieves outstanding performances on ActivityNet Captions dataset in most metrics.

We study how to leverage off-the-shelf visual and linguistic data to cope with out-of-vocabulary answers in visual question answering task. Existing large-scale visual datasets with annotations such as image class labels, bounding boxes and region descriptions are good sources for learning rich and diverse visual concepts. However, it is not straightforward how the visual concepts can be captured and transferred to visual question answering models due to missing link between question dependent answering models and visual data without question. We tackle this problem in two steps: 1) learning a task conditional visual classifier, which is capable of solving diverse question-specific visual recognition tasks, based on unsupervised task discovery and 2) transferring the task conditional visual classifier to visual question answering models. Specifically, we employ linguistic knowledge sources such as structured lexical database (e.g. WordNet) and visual descriptions for unsupervised task discovery, and transfer a learned task conditional visual classifier as an answering unit in a visual question answering model. We empirically show that the proposed algorithm generalizes to out-of-vocabulary answers successfully using the knowledge transferred from the visual dataset.

Visual Question Answering (VQA) is a notoriously challenging problem because it involves various heterogeneous tasks defined by questions within a unified framework. Learning specialized models for individual types of tasks is intuitively attracting but surprisingly difficult; it is not straightforward to outperform naive independent ensemble approach. We present a principled algorithm to learn specialized models with knowledge distillation under a multiple choice learning (MCL) framework, where training examples are assigned dynamically to a subset of models for updating network parameters. The assigned and non-assigned models are learned to predict ground-truth answers and imitate their own base models before specialization, respectively. Our approach alleviates the limitation of data deficiency in existing MCL frameworks, and allows each model to learn its own specialized expertise without forgetting general knowledge. The proposed framework is model-agnostic and applicable to any tasks other than VQA, e.g., image classification with a large number of labels but few per-class examples, which is known to be difficult under existing MCL schemes. Our experimental results indeed demonstrate that our method outperforms other baselines for VQA and image classification.

Visual attention plays an important role to understand images and demonstrates its effectiveness in generating natural language descriptions of images. On the other hand, recent studies show that language associated with an image can steer visual attention in the scene during our cognitive process. Inspired by this, we introduce a text-guided attention model for image captioning, which learns to drive visual attention using associated captions. For this model, we propose an exemplar-based learning approach that retrieves from training data associated captions with each image, and use them to learn attention on visual features. Our attention model enables to describe a detailed state of scenes by distinguishing small or confusable objects effectively. We validate our model on MS-COCO Captioning benchmark and achieve the state-of-the-art performance in standard metrics.

We present a framework to analyze various aspects of models for video question answering (VideoQA) using customizable synthetic datasets, which are constructed automatically from gameplay videos. Our work is motivated by the fact that existing models are often tested only on datasets that require excessively high-level reasoning or mostly contain instances accessible through single frame inferences. Hence, it is difficult to measure capacity and flexibility of trained models, and existing techniques often rely on adhoc implementations of deep neural networks without clear insight into datasets and models. We are particularly interested in understanding temporal relationships between video events to solve VideoQA problems; this is because reasoning temporal dependency is one of the most distinct components in videos from images. To address this objective, we automatically generate a customized synthetic VideoQA dataset using Super Mario Bros. gameplay videos so that it contains events with different levels of reasoning complexity. Using the dataset, we show that properly constructed datasets with events in various complexity levels are critical to learn effective models and improve overall performance.

Overfitting is one of the most critical challenges in deep neural networks, and there are various types of regularization methods to improve generalization performance. Injecting noises to hidden units during training, e.g., dropout, is known as a successful regularizer, but it is still not clear enough why such training techniques work well in practice and how we can maximize their benefit in the presence of two conflicting objectives---optimizing to true data distribution and preventing overfitting by regularization. This paper addresses the above issues by 1) interpreting that the conventional training methods with regularization by noise injection optimize the lower bound of the true objective and 2) proposing a technique to achieve a tighter lower bound using multiple noise samples per training example in a stochastic gradient descent iteration. We demonstrate the effectiveness of our idea in several computer vision applications.