Chong Luo

Abdelrahman Abouelenin, Atabak Ashfaq, Adam Atkinson et al. · microsoft-research

We introduce Phi-4-Mini and Phi-4-Multimodal, compact yet highly capable language and multimodal models. Phi-4-Mini is a 3.8-billion-parameter language model trained on high-quality web and synthetic data, significantly outperforming recent open-source models of similar size and matching the performance of models twice its size on math and coding tasks requiring complex reasoning. This achievement is driven by a carefully curated synthetic data recipe emphasizing high-quality math and coding datasets. Compared to its predecessor, Phi-3.5-Mini, Phi-4-Mini features an expanded vocabulary size of 200K tokens to better support multilingual applications, as well as group query attention for more efficient long-sequence generation. Phi-4-Multimodal is a multimodal model that integrates text, vision, and speech/audio input modalities into a single model. Its novel modality extension approach leverages LoRA adapters and modality-specific routers to allow multiple inference modes combining various modalities without interference. For example, it now ranks first in the OpenASR leaderboard to date, although the LoRA component of the speech/audio modality has just 460 million parameters. Phi-4-Multimodal supports scenarios involving (vision + language), (vision + speech), and (speech/audio) inputs, outperforming larger vision-language and speech-language models on a wide range of tasks. Additionally, we experiment to further train Phi-4-Mini to enhance its reasoning capabilities. Despite its compact 3.8-billion-parameter size, this experimental version achieves reasoning performance on par with or surpassing significantly larger models, including DeepSeek-R1-Distill-Qwen-7B and DeepSeek-R1-Distill-Llama-8B.

Wenming Weng, Ruoyu Feng, Yanhui Wang et al.

We present ART$\boldsymbol{\cdot}$V, an efficient framework for auto-regressive video generation with diffusion models. Unlike existing methods that generate entire videos in one-shot, ART$\boldsymbol{\cdot}$V generates a single frame at a time, conditioned on the previous ones. The framework offers three distinct advantages. First, it only learns simple continual motions between adjacent frames, therefore avoiding modeling complex long-range motions that require huge training data. Second, it preserves the high-fidelity generation ability of the pre-trained image diffusion models by making only minimal network modifications. Third, it can generate arbitrarily long videos conditioned on a variety of prompts such as text, image or their combinations, making it highly versatile and flexible. To combat the common drifting issue in AR models, we propose masked diffusion model which implicitly learns which information can be drawn from reference images rather than network predictions, in order to reduce the risk of generating inconsistent appearances that cause drifting. Moreover, we further enhance generation coherence by conditioning it on the initial frame, which typically contains minimal noise. This is particularly useful for long video generation. When trained for only two weeks on four GPUs, ART$\boldsymbol{\cdot}$V already can generate videos with natural motions, rich details and a high level of aesthetic quality. Besides, it enables various appealing applications, e.g., composing a long video from multiple text prompts.

Yanhui Wang, Jianmin Bao, Wenming Weng et al.

We present MicroCinema, a straightforward yet effective framework for high-quality and coherent text-to-video generation. Unlike existing approaches that align text prompts with video directly, MicroCinema introduces a Divide-and-Conquer strategy which divides the text-to-video into a two-stage process: text-to-image generation and image\&text-to-video generation. This strategy offers two significant advantages. a) It allows us to take full advantage of the recent advances in text-to-image models, such as Stable Diffusion, Midjourney, and DALLE, to generate photorealistic and highly detailed images. b) Leveraging the generated image, the model can allocate less focus to fine-grained appearance details, prioritizing the efficient learning of motion dynamics. To implement this strategy effectively, we introduce two core designs. First, we propose the Appearance Injection Network, enhancing the preservation of the appearance of the given image. Second, we introduce the Appearance Noise Prior, a novel mechanism aimed at maintaining the capabilities of pre-trained 2D diffusion models. These design elements empower MicroCinema to generate high-quality videos with precise motion, guided by the provided text prompts. Extensive experiments demonstrate the superiority of the proposed framework. Concretely, MicroCinema achieves SOTA zero-shot FVD of 342.86 on UCF-101 and 377.40 on MSR-VTT. See https://wangyanhui666.github.io/MicroCinema.github.io/ for video samples.

Miaosen Zhang, Qi Dai, Yifan Yang et al.

LMMs have shown impressive visual understanding capabilities, with the potential to be applied in agents, which demand strong reasoning and planning abilities. Nevertheless, existing benchmarks mostly assess their reasoning abilities in language part, where the chain-of-thought is entirely composed of text.We consider the scenario where visual signals are continuously updated and required along the decision making process. Such vision-in-the-chain reasoning paradigm is more aligned with the needs of multimodal agents, while being rarely evaluated. In this paper, we introduce MageBench, a reasoning capability oriented multimodal agent benchmark that, while having light-weight environments, poses significant reasoning challenges and holds substantial practical value. This benchmark currently includes three types of environments: WebUI, Sokoban, and Football, comprising a total of 483 different scenarios. It thoroughly validates the agent's knowledge and engineering capabilities, visual intelligence, and interaction skills. The results show that only a few product-level models are better than random acting, and all of them are far inferior to human-level. More specifically, we found current models severely lack the ability to modify their planning based on visual feedback, as well as visual imagination, interleaved image-text long context handling, and other abilities. We hope that our work will provide optimization directions for LMM from the perspective of being an agent. We release our code and data at https://github.com/microsoft/MageBench.

Rui Tian, Qi Dai, Jianmin Bao et al.

Commercial video generation models have exhibited realistic, high-fidelity results but are still restricted to limited access. One crucial obstacle for large-scale applications is the expensive training and inference cost. In this paper, we argue that videos contain significantly more redundant information than images, allowing them to be encoded with very few motion latents. Towards this goal, we design an image-conditioned VAE that projects videos into extremely compressed latent space and decode them based on content images. This magic Reducio charm enables 64x reduction of latents compared to a common 2D VAE, without sacrificing the quality. Building upon Reducio-VAE, we can train diffusion models for high-resolution video generation efficiently. Specifically, we adopt a two-stage generation paradigm, first generating a condition image via text-to-image generation, followed by text-image-to-video generation with the proposed Reducio-DiT. Extensive experiments show that our model achieves strong performance in evaluation. More importantly, our method significantly boosts the training and inference efficiency of video LDMs. Reducio-DiT is trained in just 3.2K A100 GPU hours in total and can generate a 16-frame 1024$\times$1024 video clip within 15.5 seconds on a single A100 GPU. Code released at https://github.com/microsoft/Reducio-VAE .

Tian Xie, Zitian Gao, Qingnan Ren et al.

Inspired by the success of DeepSeek-R1, we explore the potential of rule-based reinforcement learning (RL) in large reasoning models. To analyze reasoning dynamics, we use synthetic logic puzzles as training data due to their controllable complexity and straightforward answer verification. We make some key technical contributions that lead to effective and stable RL training: a system prompt that emphasizes the thinking and answering process, a stringent format reward function that penalizes outputs for taking shortcuts, and a straightforward training recipe that achieves stable convergence. Our 7B model develops advanced reasoning skills-such as reflection, verification, and summarization-that are absent from the logic corpus. Remarkably, after training on just 5K logic problems, it demonstrates generalization abilities to the challenging math benchmarks AIME and AMC.

Yucheng Zhao, Guangting Wang, Chuanxin Tang et al.

Convolutional neural networks (CNN) are the dominant deep neural network (DNN) architecture for computer vision. Recently, Transformer and multi-layer perceptron (MLP)-based models, such as Vision Transformer and MLP-Mixer, started to lead new trends as they showed promising results in the ImageNet classification task. In this paper, we conduct empirical studies on these DNN structures and try to understand their respective pros and cons. To ensure a fair comparison, we first develop a unified framework called SPACH which adopts separate modules for spatial and channel processing. Our experiments under the SPACH framework reveal that all structures can achieve competitive performance at a moderate scale. However, they demonstrate distinctive behaviors when the network size scales up. Based on our findings, we propose two hybrid models using convolution and Transformer modules. The resulting Hybrid-MS-S+ model achieves 83.9% top-1 accuracy with 63M parameters and 12.3G FLOPS. It is already on par with the SOTA models with sophisticated designs. The code and models are publicly available at https://github.com/microsoft/SPACH.

Guangting Wang, Yizhou Zhou, Chong Luo et al.

Inspired by the fact that human eyes continue to develop tracking ability in early and middle childhood, we propose to use tracking as a proxy task for a computer vision system to learn the visual representations. Modelled on the Catch game played by the children, we design a Catch-the-Patch (CtP) game for a 3D-CNN model to learn visual representations that would help with video-related tasks. In the proposed pretraining framework, we cut an image patch from a given video and let it scale and move according to a pre-set trajectory. The proxy task is to estimate the position and size of the image patch in a sequence of video frames, given only the target bounding box in the first frame. We discover that using multiple image patches simultaneously brings clear benefits. We further increase the difficulty of the game by randomly making patches invisible. Extensive experiments on mainstream benchmarks demonstrate the superior performance of CtP against other video pretraining methods. In addition, CtP-pretrained features are less sensitive to domain gaps than those trained by a supervised action recognition task. When both trained on Kinetics-400, we are pleasantly surprised to find that CtP-pretrained representation achieves much higher action classification accuracy than its fully supervised counterpart on Something-Something dataset. Code is available online: github.com/microsoft/CtP.

Shuyuan Tu, Zhen Xing, Xintong Han et al.

Current diffusion models for human image animation struggle to ensure identity (ID) consistency. This paper presents StableAnimator, the first end-to-end ID-preserving video diffusion framework, which synthesizes high-quality videos without any post-processing, conditioned on a reference image and a sequence of poses. Building upon a video diffusion model, StableAnimator contains carefully designed modules for both training and inference striving for identity consistency. In particular, StableAnimator begins by computing image and face embeddings with off-the-shelf extractors, respectively and face embeddings are further refined by interacting with image embeddings using a global content-aware Face Encoder. Then, StableAnimator introduces a novel distribution-aware ID Adapter that prevents interference caused by temporal layers while preserving ID via alignment. During inference, we propose a novel Hamilton-Jacobi-Bellman (HJB) equation-based optimization to further enhance the face quality. We demonstrate that solving the HJB equation can be integrated into the diffusion denoising process, and the resulting solution constrains the denoising path and thus benefits ID preservation. Experiments on multiple benchmarks show the effectiveness of StableAnimator both qualitatively and quantitatively.

Sunghwan Hong, Jaewoo Jung, Heeseong Shin et al.

This work delves into the task of pose-free novel view synthesis from stereo pairs, a challenging and pioneering task in 3D vision. Our innovative framework, unlike any before, seamlessly integrates 2D correspondence matching, camera pose estimation, and NeRF rendering, fostering a synergistic enhancement of these tasks. We achieve this through designing an architecture that utilizes a shared representation, which serves as a foundation for enhanced 3D geometry understanding. Capitalizing on the inherent interplay between the tasks, our unified framework is trained end-to-end with the proposed training strategy to improve overall model accuracy. Through extensive evaluations across diverse indoor and outdoor scenes from two real-world datasets, we demonstrate that our approach achieves substantial improvement over previous methodologies, especially in scenarios characterized by extreme viewpoint changes and the absence of accurate camera poses.

Yu Zhang, Yunqi Li, Yifan Yang et al.

Although chain-of-thought reasoning and reinforcement learning (RL) have driven breakthroughs in NLP, their integration into generative vision models remains underexplored. We introduce ReasonGen-R1, a two-stage framework that first imbues an autoregressive image generator with explicit text-based "thinking" skills via supervised fine-tuning on a newly generated reasoning dataset of written rationales, and then refines its outputs using Group Relative Policy Optimization. To enable the model to reason through text before generating images, We automatically generate and release a corpus of model crafted rationales paired with visual prompts, enabling controlled planning of object layouts, styles, and scene compositions. Our GRPO algorithm uses reward signals from a pretrained vision language model to assess overall visual quality, optimizing the policy in each update. Evaluations on GenEval, DPG, and the T2I benchmark demonstrate that ReasonGen-R1 consistently outperforms strong baselines and prior state-of-the-art models. More: aka.ms/reasongen.

Danlong Yuan, Tian Xie, Shaohan Huang et al.

Large reasoning models, such as OpenAI o1 or DeepSeek R1, have demonstrated remarkable performance on reasoning tasks but often incur a long reasoning path with significant memory and time costs. Existing methods primarily aim to shorten reasoning paths by introducing additional training data and stages. In this paper, we propose three critical reward designs integrated directly into the reinforcement learning process of large reasoning models, which reduce the response length without extra training stages. Experiments on four settings show that our method significantly decreases response length while maintaining or even improving performance. Specifically, in a logic reasoning setting, we achieve a 40% reduction in response length averaged by steps alongside a 14% gain in performance. For math problems, we reduce response length averaged by steps by 33% while preserving performance.

Shuyuan Tu, Zhen Xing, Xintong Han et al.

Current diffusion models for human image animation often struggle to maintain identity (ID) consistency, especially when the reference image and driving video differ significantly in body size or position. We introduce StableAnimator++, the first ID-preserving video diffusion framework with learnable pose alignment, capable of generating high-quality videos conditioned on a reference image and a pose sequence without any post-processing. Building upon a video diffusion model, StableAnimator++ contains carefully designed modules for both training and inference, striving for identity consistency. In particular, StableAnimator++ first uses learnable layers to predict the similarity transformation matrices between the reference image and the driven poses via injecting guidance from Singular Value Decomposition (SVD). These matrices align the driven poses with the reference image, mitigating misalignment to a great extent. StableAnimator++ then computes image and face embeddings using off-the-shelf encoders, refining the face embeddings via a global content-aware Face Encoder. To further maintain ID, we introduce a distribution-aware ID Adapter that counteracts interference caused by temporal layers while preserving ID via distribution alignment. During the inference stage, we propose a novel Hamilton-Jacobi-Bellman (HJB) based face optimization integrated into the denoising process, guiding the diffusion trajectory for enhanced facial fidelity. Experiments on benchmarks show the effectiveness of StableAnimator++ both qualitatively and quantitatively.

Ziqiang Xu, Qi Dai, Tian Xie et al.

Video understanding is inherently intention-driven-humans naturally focus on relevant frames based on their goals. Recent advancements in multimodal large language models (MLLMs) have enabled flexible query-driven reasoning; however, video-based frameworks like Video Chain-of-Thought lack direct training signals to effectively identify relevant frames. Current approaches often rely on heuristic methods or pseudo-label supervised annotations, which are both costly and limited in scalability across diverse scenarios. To overcome these challenges, we introduce ViaRL, the first framework to leverage rule-based reinforcement learning (RL) for optimizing frame selection in intention-driven video understanding. An iterated amplification strategy is adopted to perform alternating cyclic training in the video CoT system, where each component undergoes iterative cycles of refinement to improve its capabilities. ViaRL utilizes the answer accuracy of a downstream model as a reward signal to train a frame selector through trial-and-error, eliminating the need for expensive annotations while closely aligning with human-like learning processes. Comprehensive experiments across multiple benchmarks, including VideoMME, LVBench, and MLVU, demonstrate that ViaRL consistently delivers superior temporal grounding performance and robust generalization across diverse video understanding tasks, highlighting its effectiveness and scalability. Notably, ViaRL achieves a nearly 15\% improvement on Needle QA, a subset of MLVU, which is required to search a specific needle within a long video and regarded as one of the most suitable benchmarks for evaluating temporal grounding.

Shuyuan Tu, Yueming Pan, Yinming Huang et al.

Current diffusion models for audio-driven avatar video generation struggle to synthesize long videos with natural audio synchronization and identity consistency. This paper presents StableAvatar, the first end-to-end video diffusion transformer that synthesizes infinite-length high-quality videos without post-processing. Conditioned on a reference image and audio, StableAvatar integrates tailored training and inference modules to enable infinite-length video generation. We observe that the main reason preventing existing models from generating long videos lies in their audio modeling. They typically rely on third-party off-the-shelf extractors to obtain audio embeddings, which are then directly injected into the diffusion model via cross-attention. Since current diffusion backbones lack any audio-related priors, this approach causes severe latent distribution error accumulation across video clips, leading the latent distribution of subsequent segments to drift away from the optimal distribution gradually. To address this, StableAvatar introduces a novel Time-step-aware Audio Adapter that prevents error accumulation via time-step-aware modulation. During inference, we propose a novel Audio Native Guidance Mechanism to further enhance the audio synchronization by leveraging the diffusion's own evolving joint audio-latent prediction as a dynamic guidance signal. To enhance the smoothness of the infinite-length videos, we introduce a Dynamic Weighted Sliding-window Strategy that fuses latent over time. Experiments on benchmarks show the effectiveness of StableAvatar both qualitatively and quantitatively.

Daneul Kim, Jingxu Zhang, Wonjoon Jin et al.

We propose to train a subject-driven customized video generation model through decoupling the subject-specific learning from temporal dynamics in zero-shot without additional tuning. A traditional method for video customization that is tuning-free often relies on large, annotated video datasets, which are computationally expensive and require extensive annotation. In contrast to the previous approach, we introduce the use of an image customization dataset directly on training video customization models, factorizing the video customization into two folds: (1) identity injection through image customization dataset and (2) temporal modeling preservation with a small set of unannotated videos through the image-to-video training method. Additionally, we employ random image token dropping with randomized image initialization during image-to-video fine-tuning to mitigate the copy-and-paste issue. To further enhance learning, we introduce stochastic switching during joint optimization of subject-specific and temporal features, mitigating catastrophic forgetting. Our method achieves strong subject consistency and scalability, outperforming existing video customization models in zero-shot settings, demonstrating the effectiveness of our framework.

Miaosen Zhang, Yixuan Wei, Zhen Xing et al.

Modern vision models are trained on very large noisy datasets. While these models acquire strong capabilities, they may not follow the user's intent to output the desired results in certain aspects, e.g., visual aesthetic, preferred style, and responsibility. In this paper, we target the realm of visual aesthetics and aim to align vision models with human aesthetic standards in a retrieval system. Advanced retrieval systems usually adopt a cascade of aesthetic models as re-rankers or filters, which are limited to low-level features like saturation and perform poorly when stylistic, cultural or knowledge contexts are involved. We find that utilizing the reasoning ability of large language models (LLMs) to rephrase the search query and extend the aesthetic expectations can make up for this shortcoming. Based on the above findings, we propose a preference-based reinforcement learning method that fine-tunes the vision models to distill the knowledge from both LLMs reasoning and the aesthetic models to better align the vision models with human aesthetics. Meanwhile, with rare benchmarks designed for evaluating retrieval systems, we leverage large multi-modality model (LMM) to evaluate the aesthetic performance with their strong abilities. As aesthetic assessment is one of the most subjective tasks, to validate the robustness of LMM, we further propose a novel dataset named HPIR to benchmark the alignment with human aesthetics. Experiments demonstrate that our method significantly enhances the aesthetic behaviors of the vision models, under several metrics. We believe the proposed algorithm can be a general practice for aligning vision models with human values.

Yizhou Zhou, Chong Luo, Xiaoyan Sun et al.

Predicting future frames of video sequences is challenging due to the complex and stochastic nature of the problem. Video prediction methods based on variational auto-encoders (VAEs) have been a great success, but they require the training data to contain multiple possible futures for an observed video sequence. This is hard to be fulfilled when videos are captured in the wild where any given observation only has a determinate future. As a result, training a vanilla VAE model with these videos inevitably causes posterior collapse. To alleviate this problem, we propose a novel VAE structure, dabbed VAE-in-VAE or VAE$^2$. The key idea is to explicitly introduce stochasticity into the VAE. We treat part of the observed video sequence as a random transition state that bridges its past and future, and maximize the likelihood of a Markov Chain over the video sequence under all possible transition states. A tractable lower bound is proposed for this intractable objective function and an end-to-end optimization algorithm is designed accordingly. VAE$^2$ can mitigate the posterior collapse problem to a large extent, as it breaks the direct dependence between future and observation and does not directly regress the determinate future provided by the training data. We carry out experiments on a large-scale dataset called Cityscapes, which contains videos collected from a number of urban cities. Results show that VAE$^2$ is capable of predicting diverse futures and is more resistant to posterior collapse than the other state-of-the-art VAE-based approaches. We believe that VAE$^2$ is also applicable to other stochastic sequence prediction problems where training data are lack of stochasticity.

Yizhou Zhou, Xiaoyan Sun, Chong Luo et al.

Despite the success in still image recognition, deep neural networks for spatiotemporal signal tasks (such as human action recognition in videos) still suffers from low efficacy and inefficiency over the past years. Recently, human experts have put more efforts into analyzing the importance of different components in 3D convolutional neural networks (3D CNNs) to design more powerful spatiotemporal learning backbones. Among many others, spatiotemporal fusion is one of the essentials. It controls how spatial and temporal signals are extracted at each layer during inference. Previous attempts usually start by ad-hoc designs that empirically combine certain convolutions and then draw conclusions based on the performance obtained by training the corresponding networks. These methods only support network-level analysis on limited number of fusion strategies. In this paper, we propose to convert the spatiotemporal fusion strategies into a probability space, which allows us to perform network-level evaluations of various fusion strategies without having to train them separately. Besides, we can also obtain fine-grained numerical information such as layer-level preference on spatiotemporal fusion within the probability space. Our approach greatly boosts the efficiency of analyzing spatiotemporal fusion. Based on the probability space, we further generate new fusion strategies which achieve the state-of-the-art performance on four well-known action recognition datasets.

Guangting Wang, Chong Luo, Xiaoyan Sun et al.

We consider the tracking problem as a special type of object detection problem, which we call instance detection. With proper initialization, a detector can be quickly converted into a tracker by learning the new instance from a single image. We find that model-agnostic meta-learning (MAML) offers a strategy to initialize the detector that satisfies our needs. We propose a principled three-step approach to build a high-performance tracker. First, pick any modern object detector trained with gradient descent. Second, conduct offline training (or initialization) with MAML. Third, perform domain adaptation using the initial frame. We follow this procedure to build two trackers, named Retina-MAML and FCOS-MAML, based on two modern detectors RetinaNet and FCOS. Evaluations on four benchmarks show that both trackers are competitive against state-of-the-art trackers. On OTB-100, Retina-MAML achieves the highest ever AUC of 0.712. On TrackingNet, FCOS-MAML ranks the first on the leader board with an AUC of 0.757 and the normalized precision of 0.822. Both trackers run in real-time at 40 FPS.

Dacheng Yin, Chong Luo, Zhiwei Xiong et al.

Time-frequency (T-F) domain masking is a mainstream approach for single-channel speech enhancement. Recently, focuses have been put to phase prediction in addition to amplitude prediction. In this paper, we propose a phase-and-harmonics-aware deep neural network (DNN), named PHASEN, for this task. Unlike previous methods that directly use a complex ideal ratio mask to supervise the DNN learning, we design a two-stream network, where amplitude stream and phase stream are dedicated to amplitude and phase prediction. We discover that the two streams should communicate with each other, and this is crucial to phase prediction. In addition, we propose frequency transformation blocks to catch long-range correlations along the frequency axis. The visualization shows that the learned transformation matrix spontaneously captures the harmonic correlation, which has been proven to be helpful for T-F spectrogram reconstruction. With these two innovations, PHASEN acquires the ability to handle detailed phase patterns and to utilize harmonic patterns, getting 1.76dB SDR improvement on AVSpeech + AudioSet dataset. It also achieves significant gains over Google's network on this dataset. On Voice Bank + DEMAND dataset, PHASEN outperforms previous methods by a large margin on four metrics.

Yizhou Zhou, Xiaoyan Sun, Chong Luo et al.

The emergence of neural architecture search (NAS) has greatly advanced the research on network design. Recent proposals such as gradient-based methods or one-shot approaches significantly boost the efficiency of NAS. In this paper, we formulate the NAS problem from a Bayesian perspective. We propose explicitly estimating the joint posterior distribution over pairs of network architecture and weights. Accordingly, a hybrid network representation is presented which enables us to leverage the Variational Dropout so that the approximation of the posterior distribution becomes fully gradient-based and highly efficient. A posterior-guided sampling method is then presented to sample architecture candidates and directly make evaluations. As a Bayesian approach, our posterior-guided NAS (PGNAS) avoids tuning a number of hyper-parameters and enables a very effective architecture sampling in posterior probability space. Interestingly, it also leads to a deeper insight into the weight sharing used in the one-shot NAS and naturally alleviates the mismatch between the sampled architecture and weights caused by the weight sharing. We validate our PGNAS method on the fundamental image classification task. Results on Cifar-10, Cifar-100 and ImageNet show that PGNAS achieves a good trade-off between precision and speed of search among NAS methods. For example, it takes 11 GPU days to search a very competitive architecture with 1.98% and 14.28% test errors on Cifar10 and Cifar100, respectively.

Guangting Wang, Chong Luo, Zhiwei Xiong et al.

The greatest challenge facing visual object tracking is the simultaneous requirements on robustness and discrimination power. In this paper, we propose a SiamFC-based tracker, named SPM-Tracker, to tackle this challenge. The basic idea is to address the two requirements in two separate matching stages. Robustness is strengthened in the coarse matching (CM) stage through generalized training while discrimination power is enhanced in the fine matching (FM) stage through a distance learning network. The two stages are connected in series as the input proposals of the FM stage are generated by the CM stage. They are also connected in parallel as the matching scores and box location refinements are fused to generate the final results. This innovative series-parallel structure takes advantage of both stages and results in superior performance. The proposed SPM-Tracker, running at 120fps on GPU, achieves an AUC of 0.687 on OTB-100 and an EAO of 0.434 on VOT-16, exceeding other real-time trackers by a notable margin.

Anfeng He, Chong Luo, Xinmei Tian et al.

Recently, Siamese network based trackers have received tremendous interest for their fast tracking speed and high performance. Despite the great success, this tracking framework still suffers from several limitations. First, it cannot properly handle large object rotation. Second, tracking gets easily distracted when the background contains salient objects. In this paper, we propose two simple yet effective mechanisms, namely angle estimation and spatial masking, to address these issues. The objective is to extract more representative features so that a better match can be obtained between the same object from different frames. The resulting tracker, named Siam-BM, not only significantly improves the tracking performance, but more importantly maintains the realtime capability. Evaluations on the VOT2017 dataset show that Siam-BM achieves an EAO of 0.335, which makes it the best-performing realtime tracker to date.

Anfeng He, Chong Luo, Xinmei Tian et al.

Observing that Semantic features learned in an image classification task and Appearance features learned in a similarity matching task complement each other, we build a twofold Siamese network, named SA-Siam, for real-time object tracking. SA-Siam is composed of a semantic branch and an appearance branch. Each branch is a similarity-learning Siamese network. An important design choice in SA-Siam is to separately train the two branches to keep the heterogeneity of the two types of features. In addition, we propose a channel attention mechanism for the semantic branch. Channel-wise weights are computed according to the channel activations around the target position. While the inherited architecture from SiamFC \cite{SiamFC} allows our tracker to operate beyond real-time, the twofold design and the attention mechanism significantly improve the tracking performance. The proposed SA-Siam outperforms all other real-time trackers by a large margin on OTB-2013/50/100 benchmarks.