Navigating complex urban environments using natural language instructions poses significant challenges for embodied agents, including noisy language instructions, ambiguous spatial references, diverse landmarks, and dynamic street scenes. Current visual navigation methods are typically limited to simulated or off-street environments, and often rely on precise goal formats, such as specific coordinates or images. This limits their effectiveness for autonomous agents like last-mile delivery robots navigating unfamiliar cities. To address these limitations, we introduce UrbanNav, a scalable framework that trains embodied agents to follow free-form language instructions in diverse urban settings. Leveraging web-scale city walking videos, we develop an scalable annotation pipeline that aligns human navigation trajectories with language instructions grounded in real-world landmarks. UrbanNav encompasses over 1,500 hours of navigation data and 3 million instruction-trajectory-landmark triplets, capturing a wide range of urban scenarios. Our model learns robust navigation policies to tackle complex urban scenarios, demonstrating superior spatial reasoning, robustness to noisy instructions, and generalization to unseen urban settings. Experimental results show that UrbanNav significantly outperforms existing methods, highlighting the potential of large-scale web video data to enable language-guided, real-world urban navigation for embodied agents.

Jiatong Ma, Longteng Guo, Yuchen Liu, Zijia Zhao, Dongze Hao, Xuanxu Lin, Jing Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Detecting 3D objects from surround-view images focuses on capturing the spatio-temporal positions of the surrounding environment, serving as a pivotal capability for vision-centric autonomous driving and robotics. While existing approaches primarily employ either dense BEV queries or sparse 3D queries, both paradigms have inherent limitations: dense queries suffer from redundant feature interactions and optimization conflicts, while sparse queries rely on high-quality initialization and struggle with error propagation in complex scenarios. To address these challenges, we proposeTAS-DAQ, a novel two-stage framework that synergizes dense and sparse query strategies. In Stage I, we generate geometry-aware coarse queries through the BEV feature providing robust initialization, thereby ensuring robust query initialization with explicit 3D priors. Stage II introduces a learnable Query Bank with temporal fusion to iteratively refine sparse queries by capturing discriminative instance features across views and frames. Moreover, considering the optimization conflicts caused by redundant query interactions in dense paradigms, we introduce adaptive query aggregation in the query bank that dynamically prioritizes high-confidence queries from BEV features, effectively addressing query error propagation while enhancing instance-level representation consistency. Extensive experiments on the nuScenes R50 benchmark demonstrate state-of-the-art performance, achieving56.9 % NDSand46.1% mAP.

Most current video MLLMs rely on uniform frame sampling and image-level encoders, resulting in inefficient data processing and limited motion awareness. To address these challenges, we introduce EMA, an Efficient Motion-Aware video MLLM that utilizes compressed video structures as inputs. We propose a motion-aware GOP (Group of Pictures) encoder that fuses spatial and motion information within a GOP unit in the compressed video stream, generating compact, informative visual tokens. By integrating fewer but denser RGB frames with more but sparser motion vectors in this native slow-fast input architecture, our approach reduces redundancy and enhances motion representation. Additionally, we introduce MotionBench, a benchmark for evaluating motion understanding across four motion types: linear, curved, rotational, and contact-based. Experimental results show that EMA achieves state-of-the-art performance on both MotionBench and popular video question answering benchmarks, while reducing inference costs. Moreover, EMA demonstrates strong scalability, as evidenced by its competitive performance on long video understanding benchmarks.

Rotary Position Embedding (RoPE) has shown strong performance in text-based Large Language Models (LLMs), but extending it to video remains a challenge due to the intricate spatiotemporal structure of video frames.Existing adaptations, such as RoPE-3D, attempt to encode spatial and temporal dimensions separately but suffer from two major limitations: positional bias in attention distribution and disruptions in video-text transitions.To overcome these issues, we propose Video Rotary Position Embedding (VRoPE), a novel positional encoding method tailored for Video-LLMs.Specifically, we introduce a more balanced encoding strategy that mitigates attention biases, ensuring a more uniform distribution of spatial focus.Additionally, our approach restructures positional indices to ensure a smooth transition between video and text tokens.Extensive experiments on different models demonstrate that VRoPE consistently outperforms previous RoPE variants, achieving significant improvements in video understanding, temporal reasoning, and retrieval tasks.Code is available at https://github.com/johncaged/VRoPE.

Existing video-language models (Video-LLMs) typically rely on concatenating visual tokens with textual inputs for joint modeling.However, this token-level alignment leads to significant inefficiency, especially when scaling to long videos with dense visual inputs.In this work, we propose a video-to-parameter efficiency paradigm named ViPE that eliminates redundant visual tokens by transforming video content into visual perceptual weights, which are directly injected into the LLM's parameters.ViPE consists of a visual injection module that compresses video features into a small set of perceptual queries using a hierarchical merge strategy, and a visual perception module that integrates the resulting representations into the LLM through a lightweight LoRA-like mechanism.ViPE achieves performance comparable to token-based baselines such as LLaVA, while reducing FLOPs by 85% and inference time by up to 65%, demonstrating a highly efficient and scalable solution for video understanding.

Most Video-Large Language Models (Video-LLMs) adopt an encoder-decoder framework, where a vision encoder extracts frame-wise features for processing by a language model. However, this approach incurs high computational costs, introduces resolution biases, and struggles to capture fine-grained multimodal interactions. To overcome these limitations, we propose ELVA, an encoder-free Video-LLM that directly models nuanced video-language interactions without relying on a vision encoder. ELVA employs token merging to construct a bottom-up hierarchical representation and incorporates a video guidance supervisor for direct spatiotemporal representation learning. Additionally, a hybrid-resolution mechanism strategically integrates high- and low-resolution frames as inputs to achieve an optimal balance between performance and efficiency. With only 7M publicly available video-text pairs, ELVA achieves performance on par with encoder-based Video-LLMs while reducing FLOPs by up to 95\% and inference latency by 92\%, offering a scalable and efficient solution for real-time video understanding.

In the era of Large Language Models (LLMs), Mixture-of-Experts (MoE) architectures offer a promising approach to managing computational costs while scaling up model parameters. Conventional MoE-based LLMs typically employ static Top-K routing, which activates a fixed and equal number of experts for each token regardless of their significance within the context. In this paper, we propose a novel Ada-K routing strategy that dynamically adjusts the number of activated experts for each token, thereby improving the balance between computational efficiency and model performance. Specifically, our strategy incorporates learnable and lightweight allocator modules that decide customized expert resource allocation tailored to the contextual needs for each token. These allocators are designed to be fully pluggable, making it broadly applicable across all mainstream MoE-based LLMs. We leverage the Proximal Policy Optimization (PPO) algorithm to facilitate an end-to-end learning process for this non-differentiable decision-making framework. Extensive evaluations on four popular baseline models demonstrate that our Ada-K routing method significantly outperforms conventional Top-K routing. Compared to Top-K, our method achieves over 25% reduction in FLOPs and more than 20% inference speedup while still improving performance across various benchmarks. Moreover, the training of Ada-K is highly efficient. Even for Mixtral-8x22B, a MoE-based LLM with more than 140B parameters, the training time is limited to 8 hours. Detailed analysis shows that harder tasks, middle layers, and content words tend to activate more experts, providing valuable insights for future adaptive MoE system designs. Both the training code and model checkpoints will be publicly available.

Video understanding is a crucial next step for multimodal large language models (MLLMs). Various benchmarks are introduced for better evaluating the MLLMs. Nevertheless, current video benchmarks are still inefficient for evaluating video models during iterative development due to the high cost of constructing datasets and the difficulty in isolating specific skills. In this paper, we propose VideoNIAH (Video Needle in A Haystack), a benchmark construction framework through synthetic video generation. VideoNIAH decouples video content from their query-responses by inserting unrelated visual 'needles' into original videos. The framework automates the generation of query-response pairs using predefined rules, minimizing manual labor. The queries focus on specific aspects of video understanding, enabling more skill-specific evaluations. The separation between video content and the queries also allow for increased video variety and evaluations across different lengths. Utilizing VideoNIAH, we compile a video benchmark, VNBench, which includes tasks such as retrieval, ordering, and counting to evaluate three key aspects of video understanding: temporal perception, chronological ordering, and spatio-temporal coherence. We conduct a comprehensive evaluation of both proprietary and open-source models, uncovering significant differences in their video understanding capabilities across various tasks. Additionally, we perform an in-depth analysis of the test results and model configurations. Based on these findings, we provide some advice for improving video MLLM training, offering valuable insights to guide future research and model development.

Goal-Oriented Vision-and-Language Navigation (VLN) aims to enable agents to navigate to specified locations and identify designated target objects following natural language instruction. This approach has gained popularity due to its close alignment with real-world scenarios. However, existing studies have predominantly focused on enhancing navigation performance, neglecting the ability to locate objects at the navigation endpoint. This oversight has resulted in a significant discrepancy between the success rates of navigation and object grounding. The challenge is compounded by the complex reasoning required by the instructions and the necessity to synthesize multiperspective images of objects, which overwhelms traditional object grounding methods. We leverage the Multi-Modal Large Language Model (MLLM) to bridge this gap, allowing agents to seek assistance from these models when struggling to locate the target object. The agent conducts a multi-stage evaluation to discern the cause of its confusion and promptly extracts and updates the most relevant information for MLLM to assess. Our method is plug-and-play and model-agnostic, facilitating integration with numerous existing VLN strategies without the need for retraining. Implementing our approach across four distinct methods has improved performance on the REVERIE and SOON datasets, demonstrating the effectiveness and generalizability of our technique.

In this paper, we propose the Vision-Audio-Language Omni-peRception pretraining model (VALOR) for multimodal understanding and generation. Unlike widely-studied vision-language pretraining models, VALOR jointly models the relationships among vision, audio, and language in an end-to-end manner. It consists of three separate encoders for single modality representations and a decoder for multimodal conditional text generation. We design two pretext tasks to pretrain the VALOR model: Multimodal Grouping Alignment (MGA) and Multimodal Grouping Captioning (MGC). MGA projects vision, language, and audio into the same common space, simultaneously building vision-language, audio-language, and audiovisual-language alignment. MGC learns to generate text tokens under conditions of vision, audio, or both. To promote vision-audio-language pretraining research, we construct a large-scale, high-quality tri-modality dataset named VALOR-1M, containing 1 million audible videos with human-annotated audiovisual captions. Extensive experiments show that VALOR can learn strong multimodal correlations and generalize to various downstream tasks (e.g., retrieval, captioning, and question answering) with different input modalities (e.g., vision-language, audio-language, and audiovisual-language). VALOR achieves new state-of-the-art performance on a series of public cross-modality benchmarks.

Building scalable vision-language models to learn from diverse, multimodal data remains an open challenge. In this paper, we introduce an Efficient Vision-languagE foundation model, namely EVE, which is one unified multimodal Transformer pre-trained solely by one unified pre-training task. Specifically, EVE encodes both vision and language within a shared Transformer network integrated with modality-aware sparse Mixture-of-Experts (MoE) modules, which capture modality-specific information by selectively switching to different experts. To unify pre-training tasks of vision and language, EVE performs masked signal modeling on image-text pairs to reconstruct masked signals, i.e., image pixels and text tokens, given visible signals. This simple yet effective pre-training objective accelerates training by 4x compared to the model pre-trained with Image-Text Contrastive and Image-Text Matching losses. Owing to the combination of the unified architecture and pre-training task, EVE is easy to scale up, enabling better downstream performance with fewer resources and faster training speed. Despite its simplicity, EVE achieves state-of-the-art performance on various vision-language downstream tasks, including visual question answering, visual reasoning, and image-text retrieval.

Referring expression segmentation (RES) aims at segmenting the foreground masks of the entities that match the descriptive natural language expression. Previous datasets and methods for classic RES task heavily rely on the prior assumption that one expression must refer to object-level targets. In this paper, we take a step further to finer-grained part-level RES task. To promote the object-level RES task towards finer-grained vision-language understanding, we put forward a new multi-granularity referring expression segmentation (MRES) task and construct an evaluation benchmark called RefCOCOm by manual annotations. By employing our automatic model-assisted data engine, we build the largest visual grounding dataset namely MRES-32M, which comprises over 32.2M high-quality masks and captions on the provided 1M images. Besides, a simple yet strong model named UniRES is designed to accomplish the unified object-level and part-level grounding task. Extensive experiments on our RefCOCOm for MRES and three datasets (i.e., RefCOCO(+/g)) for classic RES task demonstrate the superiority of our method over previous state-of-the-art methods. To foster future research into fine-grained visual grounding, our benchmark RefCOCOm, the MRES-32M dataset and model UniRES will be publicly available at https://github.com/Rubics-Xuan/MRES.

Recent trends in Large Vision Language Models (LVLMs) research have been increasingly focusing on ad-vancing beyond general image understanding towards more nuanced, object-level referential comprehension. In this paper, we present and delve into the self-consistency ca-pability of LVLMs, a crucial aspect that reflects the mod-els' ability to both generate informative captions for spe-cific objects and subsequently utilize these captions to ac-curately re-identify the objects in a closed-loop process. This capability significantly mirrors the precision and reli-ability of fine- grained visual-language understanding. Our findings reveal that the self-consistency level of existing LVLMs falls short of expectations, posing limitations on their practical applicability and potential. To address this gap, we introduce a novel fine-tuning paradigm named Self-Consistency Tuning (SC-Tune). It features the syn-ergistic learning of a cyclic describer-locator system. This paradigm is not only data-efficient but also exhibits gener-alizability across multiple LVLMs. Through extensive ex-periments, we demonstrate that SC- Tune significantly ele-vates performance across a spectrum of object-level vision-language benchmarks and maintains competitive or im-proved performance on image-level vision-language bench-marks. Both our model and code will be publicly available at https://github.com/ivattyue/SC-Tune.

While large visual-language models (LVLM) have shown promising results on traditional visual question answering benchmarks, it is still challenging for them to answer complex VQA problems which requires diverse world knowledge.Motivated by the research of retrievalaugmented generation in the field of natural language processing, we use Dense Passage Retrieval (DPR) to retrieve related knowledge to help the model answer questions.However, DPR conduct retrieving in natural language space, which may not ensure comprehensive acquisition of image information.Thus, the retrieved knowledge is not truly conducive to helping answer the question, affecting the performance of the overall system.To address this issue, we propose a novel framework that leverages the visual-language model to select the key knowledge retrieved by DPR and answer questions.The framework consists of two modules: Selector and Answerer, where both are initialized by the LVLM and parameterefficiently finetuned by self-bootstrapping: find key knowledge in the retrieved knowledge documents using the Selector, and then use them to finetune the Answerer to predict answers; obtain the pseudo-labels of key knowledge documents based on the predictions of the Answerer and weak supervision labels, and then finetune the Selector to select key knowledge; repeat.Our framework significantly enhances the performance of the baseline on the challenging open-domain Knowledge-based VQA benchmark, OK-VQA, achieving a state-ofthe-art accuracy of 62.83%.Our code is publicly available at https://github.com/ haodongze/Self-KSel-QAns.

Recently, large vision language models (LVLMs) have advanced AI by integrating visual and linguistic data for tasks like visual conversation, image captioning, and visual question answering. Current LVLM research either scales up model size for performance or reduces parameters for limited computational resources. We believe both large and tiny models have unique strengths and that collaborative training yields better results than independent training. We propose Collaborative Training of Tiny-Large Vision Language Models (CTVLMs), a framework connecting large and tiny models via a projection layer and leveraging a synergistic training strategy. Our framework improves training efficiency by strengthening the interconnection between large and tiny models. Using the parameter efficiency of tiny models, we effectively align image-text features, then apply knowledge distillation to help large models better align cross-modal information. During fine-tuning, the large model's extensive knowledge enhances tiny model's performance. This collaborative approach allows models to adapt to various computational resources and outperforms existing methods in vision-language tasks.

Sounding Video Generation (SVG) is an audio-video joint generation task challenged by high-dimensional signal spaces, distinct data formats, and different patterns of content information. To address these issues, we introduce a novel multi-modal latent diffusion model (MM-LDM) for the SVG task. We first unify the representation of audio and video data by converting them into a single or a couple of images. Then, we introduce a hierarchical multi-modal autoencoder that constructs a low-level perceptual latent space for each modality and a shared high-level semantic feature space. The former space is perceptually equivalent to the raw signal space of each modality but drastically reduces signal dimensions. The latter space serves to bridge the information gap between modalities and provides more insightful cross-modal guidance. Our proposed method achieves new state-of-the-art results with significant quality and efficiency gains. Specifically, our method achieves a comprehensive improvement on all evaluation metrics and a faster training and sampling speed on Landscape and AIST++ datasets. Moreover, we explore its performance on open-domain sounding video generation, long sounding video generation, audio continuation, video continuation, and conditional single-modal generation tasks for a comprehensive evaluation, where our MM-LDM demonstrates exciting adaptability and generalization ability.

Referring image segmentation (RIS) is a fundamental vision-language task that intends to segment a desired object from an image based on a given natural language expression. Due to the essentially distinct data properties between image and text, most of existing methods either introduce complex designs towards fine-grained vision-language alignment or lack required dense alignment, resulting in scalability issues or mis-segmentation problems such as over- or under-segmentation. To achieve effective and efficient fine-grained feature alignment in the RIS task, we explore the potential of masked multimodal modeling coupled with self-distillation and propose a novel cross-modality masked self-distillation framework named CM-MaskSD, in which our method inherits the transferred knowledge of image-text semantic alignment from CLIP model to realize fine-grained patch-word feature alignment for better segmentation accuracy. Moreover, our CM-MaskSD framework can considerably boost model performance in a nearly parameter-free manner, since it shares weights between the main segmentation branch and the introduced masked self-distillation branches, and solely introduces negligible parameters for coordinating the multimodal features. Comprehensive experiments on three benchmark datasets (ie RefCOCO, RefCOCO+, G-Ref) for the RIS task convincingly demonstrate the superiority of our proposed framework over previous state-of-the-art methods.

Large pre-trained multimodal models have demonstrated significant success in a range of downstream tasks, including image captioning, image-text retrieval, visual question answering (VQA), etc. However, many of these methods rely on image-text pairs collected from the web as pre-training data and unfortunately overlook the need for fine-grained feature alignment between vision and language modalities, which requires detailed understanding of images and language expressions. While integrating VQA and dense captioning (DC) into pre-training can address this issue, acquiring image-question-answer as well as image-location-caption triplets is challenging and time-consuming. Additionally, publicly available datasets for VQA and dense captioning are typically limited in scale due to manual data collection and labeling efforts. In this paper, we propose a novel method called Joint QA and DC GEneration (JADE), which utilizes a pre-trained multimodal model and easily-crawled image-text pairs to automatically generate and filter large-scale VQA and dense captioning datasets. We apply this method to the Conceptual Caption (CC3M) dataset to generate a new dataset called CC3M-QA-DC. Experiments show that when used for pre-training in a multi-task manner, CC3M-QA-DC can improve the performance with various backbones on various downstream tasks. Furthermore, our generated CC3M-QA-DC can be combined with larger image-text datasets (e.g., CC15M) and achieve competitive results compared with models using much more data. Code and dataset are available at https://github.com/johncaged/OPT_Questioner.

Multimodal representation learning has shown promising improvements on various vision-language tasks (e.g., image-text retrieval, visual question answering, etc) and has significantly advanced the development of multimedia information systems. Most existing methods excel at building global-level alignment between vision and language while lacking effective fine-grained image-text interaction. In this paper, we propose a jointly masked multimodal modeling method to learn fine-grained multimodal representations. Our method performs joint masking on image-text input and integrates both implicit and explicit targets for the masked signals to recover. The implicit target provides a unified and debiased objective for vision and language, where the model predicts latent multimodal representations of the unmasked input. The explicit target further enriches the multimodal representations by recovering high-level and semantically meaningful information: momentum visual features of image patches and concepts of word tokens. Through such a masked modeling process, our model not only learns fine-grained multimodal interaction, but also avoids the semantic gap between high-level representations and low-or mid-level prediction targets (e.g., image pixels, discrete vision tokens), thus producing semantically rich multimodal representations that perform well on both zero-shot and fine-tuned settings. Our pre-trained model (named MAMO) achieves state-of-the-art performance on various downstream vision-language tasks, including image-text retrieval, visual question answering, visual reasoning, and weakly-supervised visual grounding.

The quality of video representation directly decides the performance of video related tasks, for both understanding and generation. In this paper, we propose single-modality pretrained feature fusion technique which is composed of reasonable multi-view feature extraction method and designed multi-modality feature fusion strategy. We conduct comprehensive ablation studies on MSR-VTT dataset to demonstrate the effectiveness of proposed method and it surpasses the state-of-the-art methods on both MSR-VTT and VATEX datasets. We further propose the multi-modality pretrained model finetuning technique and dataset augmentation scheme to improve the model's generalization capability. Based on these two proposed pretraining techniques and dataset augmentation scheme, we win the first place in the video captioning track of the MM21 pretraining for video understanding challenge.

Self-attention (SA) network has shown profound value in image captioning. In this paper, we improve SA from two aspects to promote the performance of image captioning. First, we propose Normalized Self-Attention (NSA), a reparameterization of SA that brings the benefits of normalization inside SA. While normalization is previously only applied outside SA, we introduce a novel normalization method and demonstrate that it is both possible and beneficial to perform it on the hidden activations inside SA. Second, to compensate for the major limit of Transformer that it fails to model the geometry structure of the input objects, we propose a class of Geometry-aware Self-Attention (GSA) that extends SA to explicitly and efficiently consider the relative geometry relations between the objects in the image. To construct our image captioning model, we combine the two modules and apply it to the vanilla self-attention network. We extensively evaluate our proposals on MS-COCO image captioning dataset and superior results are achieved when comparing to state-of-the-art approaches. Further experiments on three challenging tasks, i.e. video captioning, machine translation, and visual question answering, show the generality of our methods.

Image captioning aims to first observe an image, most notably the involved objects that are highly context-dependent, and then depict it with a natural description. However, most of the current models solely use the isolated objects vectors as image representations, ignoring the contexts among them. In this paper, we introduce a Local-Global Context (LGC) network, endowing the independent object features with shortrange perception (local contexts) and long-range dependence (global contexts). LGC network can be viewed as feature refiner, much beneficial to reason the novel objects and verbal words for the caption decoder. The local contexts are modeled with 1-D group convolution on adjacent objects, strengthening the local connections. Still further, self-attention mechanism is utilized to model the global contexts by correlating all the local contexts. Extensive experiments on MSCOCO dataset demonstrate that LGC network can easily plug into almost any neural captioning models and significantly improve the model performance.

Most image captioning models are autoregressive, i.e. they generate each word by conditioning on previously generated words, which leads to heavy latency during inference. Recently, non-autoregressive decoding has been proposed in machine translation to speed up the inference time by generating all words in parallel. Typically, these models use the word-level cross-entropy loss to optimize each word independently. However, such a learning process fails to consider the sentence-level consistency, thus resulting in inferior generation quality of these non-autoregressive models. In this paper, we propose a Non-Autoregressive Image Captioning (NAIC) model with a novel training paradigm: Counterfactuals-critical Multi-Agent Learning (CMAL). CMAL formulates NAIC as a multi-agent reinforcement learning system where positions in the target sequence are viewed as agents that learn to cooperatively maximize a sentence-level reward. Besides, we propose to utilize massive unlabeled images to boost captioning performance. Extensive experiments on MSCOCO image captioning benchmark show that our NAIC model achieves a performance comparable to state-of-the-art autoregressive models, while brings 13.9x decoding speedup.

The encoder-decoder framework has been the base of popular image captioning models, which typically predicts the target sentence based on the encoded source image one word at a time in sequence. However, such a single-pass decoding framework encounters two problems. First, mistakes in the predicted words cannot be corrected and may propagate to the entire sentence. Second, because the single-pass decoder cannot access the following un-generated words, it can only perform local planning to choose every single word according to the preceding words, while lacks the global planning ability as for maintaining the semantic consistency and fluency of the whole sentence. In order to address the above two problems, in this work, we design a ruminant captioning framework which contains an image encoder, a base decoder, and a ruminant decoder. Specifically, the outputs of the former/base decoder are utilized as the global information to guide the words prediction of the latter/ruminant decoder, in an attempt to mimic human polishing process. We enable jointly training of the whole framework and overcome the non-differential problem of discrete words by designing a novel reinforcement learning based optimization algorithm. Experiments on two datasets (MS COCO and Flickr30 k) demonstrate that our ruminant decoding method can bring significant improvements over traditional single-pass decoding based models and achieves state-of-the-art performance.

In this paper, we propose an adversarial learning network for the task of multi-style image captioning (MSCap) with a standard factual image caption dataset and a multi-stylized language corpus without paired images. How to learn a single model for multi-stylized image captioning with unpaired data is a challenging and necessary task, whereas rarely studied in previous works. The proposed framework mainly includes four contributive modules following a typical image encoder. First, a style dependent caption generator to output a sentence conditioned on an encoded image and a specified style. Second, a caption discriminator is presented to distinguish the input sentence to be real or not. The discriminator and the generator are trained in an adversarial manner to enable more natural and human-like captions. Third, a style classifier is employed to discriminate the specific style of the input sentence. Besides, a back-translation module is designed to enforce the generated stylized captions are visually grounded, with the intuition of the cycle consistency for factual caption and stylized caption. We enable an end-to-end optimization of the whole model with differentiable softmax approximation. At last, we conduct comprehensive experiments using a combined dataset containing four caption styles to demonstrate the outstanding performance of our proposed method.

Image captioning attempts to generate a sentence composed of several linguistic words, which are used to describe objects, attributes, and interactions in an image, denoted as visual semantic units in this paper. Based on this view, we propose to explicitly model the object interactions in semantics and geometry based on Graph Convolutional Networks (GCNs), and fully exploit the alignment between linguistic words and visual semantic units for image captioning. Particularly, we construct a semantic graph and a geometry graph, where each node corresponds to a visual semantic unit, i.e., an object, an attribute, or a semantic (geometrical) interaction between two objects. Accordingly, the semantic (geometrical) context-aware embeddings for each unit are obtained through the corresponding GCN learning processers. At each time step, a context gated attention module takes as inputs the embeddings of the visual semantic units and hierarchically align the current word with these units by first deciding which type of visual semantic unit (object, attribute, or interaction) the current word is about, and then finding the most correlated visual semantic units under this type. Extensive experiments are conducted on the challenging MS-COCO image captioning dataset, and superior results are reported when comparing to state-of-the-art approaches.

For sketch-based image retrieval (SBIR), we propose a generative adversarial network trained on a large number of sketches and their corresponding real images. To imitate human search process, we attempt to match candidate images with theimaginary image in user single s mind instead of the sketch query, i.e., not only the shape information of sketches but their possible content information are considered in SBIR. Specifically, a conditional generative adversarial network (cGAN) is employed to enrich the content information of sketches and recover the imaginary images, and two VGG-based encoders, which work on real and imaginary images respectively, are used to constrain their perceptual consistency from the view of feature representations. During SBIR, we first generate an imaginary image from a given sketch via cGAN, and then take the output of the learned encoder for imaginary images as the feature of the query sketch. Finally, we build an interactive SBIR system that shows encouraging performance.