Fan Yang

Shitao Xiao, Zheng Liu, Weihao Han et al. · microsoft-research

Vector quantization (VQ) based ANN indexes, such as Inverted File System (IVF) and Product Quantization (PQ), have been widely applied to embedding based document retrieval thanks to the competitive time and memory efficiency. Originally, VQ is learned to minimize the reconstruction loss, i.e., the distortions between the original dense embeddings and the reconstructed embeddings after quantization. Unfortunately, such an objective is inconsistent with the goal of selecting ground-truth documents for the input query, which may cause severe loss of retrieval quality. Recent works identify such a defect, and propose to minimize the retrieval loss through contrastive learning. However, these methods intensively rely on queries with ground-truth documents, whose performance is limited by the insufficiency of labeled data. In this paper, we propose Distill-VQ, which unifies the learning of IVF and PQ within a knowledge distillation framework. In Distill-VQ, the dense embeddings are leveraged as "teachers", which predict the query's relevance to the sampled documents. The VQ modules are treated as the "students", which are learned to reproduce the predicted relevance, such that the reconstructed embeddings may fully preserve the retrieval result of the dense embeddings. By doing so, Distill-VQ is able to derive substantial training signals from the massive unlabeled data, which significantly contributes to the retrieval quality. We perform comprehensive explorations for the optimal conduct of knowledge distillation, which may provide useful insights for the learning of VQ based ANN index. We also experimentally show that the labeled data is no longer a necessity for high-quality vector quantization, which indicates Distill-VQ's strong applicability in practice.

Yufei Zhan, Yousong Zhu, Zhiyang Chen et al.

Replicating the innate human ability to detect all objects based on free-form texts at any granularity remains a formidable challenge for Large Vision Language Models (LVLMs). Current LVLMs are predominantly constrained to locate a single, pre-existing object. This limitation leads to a compromise in model design, necessitating the introduction of visual expert models or customized head structures. Beyond these constraints, our research uncovers LVLMs' capability for basic object perception, allowing them to accurately identify and locate objects of interest. Building on this insight, we introduce a novel Language-prompted Localization Dataset to fully unleash the capabilities of LVLMs in fine-grained object perception and precise location awareness. More importantly, we present Griffon, a purely LVLM-based baseline, which does not introduce any special tokens, expert models, or additional detection modules. It simply maintains a consistent structure with popular LVLMs by unifying data formats across various localization-related scenarios and is trained end-to-end through a well-designed pipeline. Comprehensive experiments demonstrate that Griffon not only achieves state-of-the-art performance on the fine-grained RefCOCO series and Flickr30K Entities but also approaches the capabilities of the expert model Faster RCNN on the detection benchmark MSCOCO. Data, codes, and models are released at https://github.com/jefferyZhan/Griffon.

Jiaxin Deng, Dong Shen, Shiyao Wang et al.

In recent years, live streaming platforms have gained immense popularity as they allow users to broadcast their videos and interact in real-time with hosts and peers. Due to the dynamic changes of live content, accurate recommendation models are crucial for enhancing user experience. However, most previous works treat the live as a whole item and explore the Click-through-Rate (CTR) prediction framework on item-level, neglecting that the dynamic changes that occur even within the same live room. In this paper, we proposed a ContentCTR model that leverages multimodal transformer for frame-level CTR prediction. First, we present an end-to-end framework that can make full use of multimodal information, including visual frames, audio, and comments, to identify the most attractive live frames. Second, to prevent the model from collapsing into a mediocre solution, a novel pairwise loss function with first-order difference constraints is proposed to utilize the contrastive information existing in the highlight and non-highlight frames. Additionally, we design a temporal text-video alignment module based on Dynamic Time Warping to eliminate noise caused by the ambiguity and non-sequential alignment of visual and textual information. We conduct extensive experiments on both real-world scenarios and public datasets, and our ContentCTR model outperforms traditional recommendation models in capturing real-time content changes. Moreover, we deploy the proposed method on our company platform, and the results of online A/B testing further validate its practical significance.

Chao Cao, Hongbiao Zhu, Fan Yang et al.

Autonomous Exploration Development Environment is an open-source repository released to facilitate the development of high-level planning algorithms and integration of complete autonomous navigation systems. The repository contains representative simulation environment models, fundamental navigation modules, e.g., local planner, terrain traversability analysis, waypoint following, and visualization tools. Together with two of our high-level planner releases -- TARE planner for exploration and FAR planner for route planning, we detail usage of the three open-source repositories and share experiences in the integration of autonomous navigation systems. We use DARPA Subterranean Challenge as a use case where the repositories together form the main navigation system of the CMU-OSU Team. In the end, we discuss a few potential use cases in extended applications.

Yufei Zhan, Hongyin Zhao, Yousong Zhu et al.

Large Multimodal Models (LMMs) have achieved significant breakthroughs in various vision-language and vision-centric tasks based on auto-regressive modeling. However, these models typically focus on either vision-centric tasks, such as visual grounding and region description, or vision-language tasks, like image caption and multi-scenario VQAs. None of the LMMs have yet comprehensively unified both types of tasks within a single model, as seen in Large Language Models in the natural language processing field. Furthermore, even with abundant multi-task instruction-following data, directly stacking these data for universal capabilities extension remains challenging. To address these issues, we introduce a novel multi-dimension curated and consolidated multimodal dataset, named CCMD-8M, which overcomes the data barriers of unifying vision-centric and vision-language tasks through multi-level data curation and multi-task consolidation. More importantly, we present Griffon-G, a general large multimodal model that addresses both vision-centric and vision-language tasks within a single end-to-end paradigm. Griffon-G resolves the training collapse issue encountered during the joint optimization of these tasks, achieving better training efficiency. Evaluations across multimodal benchmarks, general Visual Question Answering (VQA) tasks, scene text-centric VQA tasks, document-related VQA tasks, Referring Expression Comprehension, and object detection demonstrate that Griffon-G surpasses the advanced LMMs and achieves expert-level performance in complicated vision-centric tasks.

Fan Yang, Mengqing Jiang, Chenxi Xu

SLAM system using only point cloud has been proven successful in recent years. In most of these systems, they extract features for tracking after ground removal, which causes large variance on the z-axis. Ground actually provides robust information to obtain [t_z, θ_{roll}, θ_{pitch}]$. In this project, we followed the LeGO-LOAM, a light-weighted real-time SLAM system that extracts and registers ground as an addition to the original LOAM, and we proposed a new clustering-based method to refine the planar extraction algorithm for ground such that the system can handle much more noisy or dynamic environments. We implemented this method and compared it with LeGo-LOAM on our collected data of CMU campus, as well as a collected dataset for ATV (All-Terrain Vehicle) for off-road self-driving. Both visualization and evaluation results show obvious improvement of our algorithm.

Fan Yang, Dung-Han Lee, John Keller et al.

Currently, state-of-the-art exploration methods maintain high-resolution map representations in order to optimize exploration goals in each step that maximizes information gain. However, during exploring, those "optimal" selections could quickly become obsolete due to the influx of new information, especially in large-scale environments, and result in high-frequency re-planning that hinders the overall exploration efficiency. In this paper, we propose a graph-based topological planning framework, building a sparse topological map in three-dimensional (3D) space to guide exploration steps with high-level intents so as to render consistent exploration maneuvers. Specifically, this work presents a novel method to estimate 3D space's geometry with convex polyhedrons. Then, the geometry information is utilized to group space into distinctive regions. And those regions are added as nodes into the topological map, directing the exploration process. We compared our method with the state-of-the-art in simulated environments. The proposed method achieves higher space coverage and outperforms exploration efficiency by more than 40% during experiments. Finally, a field experiment was conducted to further evaluate the applicability of our method to empower efficient and robust exploration in real-world environments.