Ting Sun

Cheng Cui, Ting Sun, Suyin Liang et al.

We introduce PaddleOCR-VL-1.5, an upgraded model achieving a new state-of-the-art (SOTA) accuracy of 94.5% on OmniDocBench v1.5. To rigorously evaluate robustness against real-world physical distortions, including scanning, skew, warping, screen-photography, and illumination, we propose the Real5-OmniDocBench benchmark. Experimental results demonstrate that this enhanced model attains SOTA performance on the newly curated benchmark. Furthermore, we extend the model's capabilities by incorporating seal recognition and text spotting tasks, while remaining a 0.9B ultra-compact VLM with high efficiency. Code: https://github.com/PaddlePaddle/PaddleOCR

Yiyou Sun, Xinyang Han, Weichen Zhang et al.

Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 subfields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact.

Zelun Zhang, Hongen Liu, Suyin Liang et al.

We introduce PaddleOCR-VL-1.6, an upgraded compact document parsing model built upon PaddleOCR-VL-1.5. Although PaddleOCR-VL-1.5 establishes a strong 0.9B baseline, its remaining errors concentrate in under-optimized regions where model behavior is unstable, data coverage is sparse, or supervision is unreliable. Rather than expanding the training corpus indiscriminately, PaddleOCR-VL-1.6 introduces a region-aware data optimization framework that identifies weak regions from the previous model, applies targeted enhancement to these regions, and improves the reliability of supervision signals. It further adopts a progressive post-training recipe based on curated data selection and reinforcement learning, pushing model performance to a higher level through staged optimization. PaddleOCR-VL-1.6 achieves a new state-of-the-art score of 96.33% on OmniDocBench v1.6, demonstrates strong competitiveness against top-tier VLMs, and provides a practical post-training recipe for the PaddleOCR-VL series.

Shiyu Tang, Ting Sun, Juncai Peng et al.

The success of transformers in computer vision has led to several attempts to adapt them for mobile devices, but their performance remains unsatisfactory in some real-world applications. To address this issue, we propose PP-MobileSeg, a semantic segmentation model that achieves state-of-the-art performance on mobile devices. PP-MobileSeg comprises three novel parts: the StrideFormer backbone, the Aggregated Attention Module (AAM), and the Valid Interpolate Module (VIM). The four-stage StrideFormer backbone is built with MV3 blocks and strided SEA attention, and it is able to extract rich semantic and detailed features with minimal parameter overhead. The AAM first filters the detailed features through semantic feature ensemble voting and then combines them with semantic features to enhance the semantic information. Furthermore, we proposed VIM to upsample the downsampled feature to the resolution of the input image. It significantly reduces model latency by only interpolating classes present in the final prediction, which is the most significant contributor to overall model latency. Extensive experiments show that PP-MobileSeg achieves a superior tradeoff between accuracy, model size, and latency compared to other methods. On the ADE20K dataset, PP-MobileSeg achieves 1.57% higher accuracy in mIoU than SeaFormer-Base with 32.9% fewer parameters and 42.3% faster acceleration on Qualcomm Snapdragon 855. Source codes are available at https://github.com/PaddlePaddle/PaddleSeg/tree/release/2.8.

Cheng Cui, Ting Sun, Suyin Liang et al.

Document parsing is a fine-grained task where image resolution significantly impacts performance. While advanced research leveraging vision-language models benefits from high-resolution input to boost model performance, this often leads to a quadratic increase in the number of vision tokens and significantly raises computational costs. We attribute this inefficiency to substantial visual regions redundancy in document images, like background. To tackle this, we propose PaddleOCR-VL, a novel coarse-to-fine architecture that focuses on semantically relevant regions while suppressing redundant ones, thereby improving both efficiency and performance. Specifically, we introduce a lightweight Valid Region Focus Module (VRFM) which leverages localization and contextual relationship prediction capabilities to identify valid vision tokens. Subsequently, we design and train a compact yet powerful 0.9B vision-language model (PaddleOCR-VL-0.9B) to perform detailed recognition, guided by VRFM outputs to avoid direct processing of the entire large image. Extensive experiments demonstrate that PaddleOCR-VL achieves state-of-the-art performance in both page-level parsing and element-level recognition. It significantly outperforms existing solutions, exhibits strong competitiveness against top-tier VLMs, and delivers fast inference while utilizing substantially fewer vision tokens and parameters, highlighting the effectiveness of targeted coarse-to-fine parsing for accurate and efficient document understanding. The source code and models are publicly available at https://github.com/PaddlePaddle/PaddleOCR.

Cheng Cui, Yubo Zhang, Ting Sun et al.

The advent of "OCR 2.0" and large-scale vision-language models (VLMs) has set new benchmarks in text recognition. However, these unified architectures often come with significant computational demands, challenges in precise text localization within complex layouts, and a propensity for textual hallucinations. Revisiting the prevailing notion that model scale is the sole path to high accuracy, this paper introduces PP-OCRv5, a meticulously optimized, lightweight OCR system with merely 5 million parameters. We demonstrate that PP-OCRv5 achieves performance competitive with many billion-parameter VLMs on standard OCR benchmarks, while offering superior localization precision and reduced hallucinations. The cornerstone of our success lies not in architectural expansion but in a data-centric investigation. We systematically dissect the role of training data by quantifying three critical dimensions: data difficulty, data accuracy, and data diversity. Our extensive experiments reveal that with a sufficient volume of high-quality, accurately labeled, and diverse data, the performance ceiling for traditional, efficient two-stage OCR pipelines is far higher than commonly assumed. This work provides compelling evidence for the viability of lightweight, specialized models in the large-model era and offers practical insights into data curation for OCR. The source code and models are publicly available at https://github.com/PaddlePaddle/PaddleOCR.

Haifeng Wang, Hua Wu, Tian Wu et al.

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Cheng Cui, Ting Sun, Manhui Lin et al.

This technical report introduces PaddleOCR 3.0, an Apache-licensed open-source toolkit for OCR and document parsing. To address the growing demand for document understanding in the era of large language models, PaddleOCR 3.0 presents three major solutions: (1) PP-OCRv5 for multilingual text recognition, (2) PP-StructureV3 for hierarchical document parsing, and (3) PP-ChatOCRv4 for key information extraction. Compared to mainstream vision-language models (VLMs), these models with fewer than 100 million parameters achieve competitive accuracy and efficiency, rivaling billion-parameter VLMs. In addition to offering a high-quality OCR model library, PaddleOCR 3.0 provides efficient tools for training, inference, and deployment, supports heterogeneous hardware acceleration, and enables developers to easily build intelligent document applications.

Ting Sun, Penghan Wang, Fan Lai

Large language models (LLMs) have facilitated a wide range of applications with distinct service-level objectives (SLOs), from latency-sensitive online tasks like interactive chatbots to throughput-oriented offline workloads like data synthesis. The existing deployment model, which dedicates machines to each workload, simplifies SLO management but often leads to poor resource utilization. This paper introduces HyGen, an interference-aware LLM serving system that enables efficient co-location of online and offline workloads while preserving SLOs. HyGen incorporates two key innovations: (1) performance control mechanisms, including a latency predictor to estimate batch execution time and an SLO-aware profiler to quantify latency interference, and (2) SLO-aware offline scheduling policies that maximize serving throughput and prevent starvation. Our evaluation on production workloads shows that HyGen achieves up to 3.9-5.8x throughput gains over online and hybrid serving baselines, while ensuring latency SLOs. The code of HyGen is publicly available at https://github.com/UIUC-MLSys/HyGen.

Cheng Cui, Ting Sun, Suyin Liang et al.

In this report, we propose PaddleOCR-VL, a SOTA and resource-efficient model tailored for document parsing. Its core component is PaddleOCR-VL-0.9B, a compact yet powerful vision-language model (VLM) that integrates a NaViT-style dynamic resolution visual encoder with the ERNIE-4.5-0.3B language model to enable accurate element recognition. This innovative model efficiently supports 109 languages and excels in recognizing complex elements (e.g., text, tables, formulas, and charts), while maintaining minimal resource consumption. Through comprehensive evaluations on widely used public benchmarks and in-house benchmarks, PaddleOCR-VL achieves SOTA performance in both page-level document parsing and element-level recognition. It significantly outperforms existing solutions, exhibits strong competitiveness against top-tier VLMs, and delivers fast inference speeds. These strengths make it highly suitable for practical deployment in real-world scenarios. Code is available at https://github.com/PaddlePaddle/PaddleOCR .

Ting Sun, Cheng Cui, Yuning Du et al.

Document layout analysis is a critical preprocessing step in document intelligence, enabling the detection and localization of structural elements such as titles, text blocks, tables, and formulas. Despite its importance, existing layout detection models face significant challenges in generalizing across diverse document types, handling complex layouts, and achieving real-time performance for large-scale data processing. To address these limitations, we present PP-DocLayout, which achieves high precision and efficiency in recognizing 23 types of layout regions across diverse document formats. To meet different needs, we offer three models of varying scales. PP-DocLayout-L is a high-precision model based on the RT-DETR-L detector, achieving 90.4% mAP@0.5 and an end-to-end inference time of 13.4 ms per page on a T4 GPU. PP-DocLayout-M is a balanced model, offering 75.2% mAP@0.5 with an inference time of 12.7 ms per page on a T4 GPU. PP-DocLayout-S is a high-efficiency model designed for resource-constrained environments and real-time applications, with an inference time of 8.1 ms per page on a T4 GPU and 14.5 ms on a CPU. This work not only advances the state of the art in document layout analysis but also provides a robust solution for constructing high-quality training data, enabling advancements in document intelligence and multimodal AI systems. Code and models are available at https://github.com/PaddlePaddle/PaddleX .

Ting Sun, Penghan Wang, Fan Lai

The rapid rise of large language models (LLMs) in text streaming services has introduced significant cost and Quality of Experience (QoE) challenges in serving millions of daily requests, especially in meeting Time-To-First-Token (TTFT) and Time-Between-Token (TBT) requirements for real-time interactions. Our real-world measurements show that both server-based and on-device deployments struggle to meet diverse QoE demands: server deployments face high costs and last-hop issues (e.g., Internet latency and dynamics), while on-device LLM inference is constrained by resources. We introduce DiSCo, a device-server cooperative scheduler designed to optimize users' QoE by adaptively routing requests and migrating response generation between endpoints while maintaining cost constraints. DiSCo employs cost-aware scheduling, leveraging the predictable speed of on-device LLM inference with the flexible capacity of server-based inference to dispatch requests on the fly, while introducing a token-level migration mechanism to ensure consistent token delivery during migration. Evaluations on real-world workloads -- including commercial services like OpenAI GPT and DeepSeek, and open-source deployments such as LLaMA3 -- show that DiSCo can improve users' QoE by reducing tail TTFT (11-52\%) and mean TTFT (6-78\%) across different model-device configurations, while dramatically reducing serving costs by up to 84\% through its migration mechanism while maintaining comparable QoE levels.

Long Phan, Alice Gatti, Ziwen Han et al. · amazon-science, apple-ml

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.

Junjie Zhang, Jingyi Xi, Zhuoyang Song et al.

Training large language models (LLMs) to act as autonomous agents for multi-turn, long-horizon tasks remains significant challenges in scalability and training efficiency. To address this, we introduce L-Zero (L0), a scalable, end-to-end training pipeline for general-purpose agents. Featuring a low-cost, extensible, and sandboxed concurrent agent worker pool, L0 lowers the barrier for applying reinforcement learning in complex environments. We also introduce NB-Agent, the agent scaffold within L0, which operates in a "code-as-action" fashion via a Read-Eval-Print-Loop (REPL). We evaluate L0 on factuality question-answering benchmarks. Our experiments demonstrate that a base model can develop robust problem-solving skills using solely Reinforcement Learning with Verifiable Rewards (RLVR). On the Qwen2.5-7B-Instruct model, our method boosts accuracy on SimpleQA from 30 % to 80 % and on HotpotQA from 22 % to 41 %. We have open-sourced the entire L0 system, including our L0 series models, the NB-Agent, a complete training pipeline, and the corresponding training recipes on (https://github.com/cmriat/l0).

Ting Sun, Junjie Zhang, Xiao Yan et al.

Modern Large Foundation Model (LFM) training has transformed the data pipeline from a static ingestion layer into a dynamic component that must co-evolve with the training process. Existing systems are ill-equipped: colocated dataloaders offer no failure isolation, while message queue-based disaggregated dataloaders operate on a record/offset abstraction that cannot express the batch-level semantics required by distributed training. We present Lakestream, a brokerless, object-store-native training data plane with three key properties. First, it introduces the Transactional Global Batch (TGB), which builds on lakehouse-style ACID storage semantics and extends them with training-specific consistency, including atomic all-rank batch visibility, a globally ordered step sequence, checkpoint-aligned lifecycle management, and end-to-end exactly-once recovery. Second, it realizes recovery and retention directly in the storage layer, by inlining producer state in the manifest and tying reclamation to distributed checkpoint state. Third, its Decentralized Adaptive Commit (DAC) algorithm sustains stable ingestion throughput as the manifest grows, without any inter-producer communication. Evaluations on large-scale multimodal pre-training and SFT workloads using 64 GPUs show that Lakestream outperforms colocated dataloader throughput while providing full failure isolation, outperforms Apache Kafka in ingestion throughput, and achieves lower consumer read latency than Kafka.

Xiaoze Liu, Ting Sun, Tianyang Xu et al.

Large Language Models (LLMs) have transformed machine learning but raised significant legal concerns due to their potential to produce text that infringes on copyrights, resulting in several high-profile lawsuits. The legal landscape is struggling to keep pace with these rapid advancements, with ongoing debates about whether generated text might plagiarize copyrighted materials. Current LLMs may infringe on copyrights or overly restrict non-copyrighted texts, leading to these challenges: (i) the need for a comprehensive evaluation benchmark to assess copyright compliance from multiple aspects; (ii) evaluating robustness against safeguard bypassing attacks; and (iii) developing effective defense targeted against the generation of copyrighted text. To tackle these challenges, we introduce a curated dataset to evaluate methods, test attack strategies, and propose lightweight, real-time defense to prevent the generation of copyrighted text, ensuring the safe and lawful use of LLMs. Our experiments demonstrate that current LLMs frequently output copyrighted text, and that jailbreaking attacks can significantly increase the volume of copyrighted output. Our proposed defense mechanism significantly reduces the volume of copyrighted text generated by LLMs by effectively refusing malicious requests. Code is publicly available at https://github.com/xz-liu/SHIELD

Chaobo Jia, Ruipeng Wan, Ting Sun et al.

LLM-based game generation promises to turn natural-language specifications into executable games, but progress is limited by the lack of reliable automated verification. Unlike conventional code generation, game correctness is defined over long-horizon interaction: a game may appear correct while violating core mechanics such as state updates, interaction rules, and phase transitions. Existing Agent-as-a-Verifier approaches collapse verification into open-ended gameplay, making verdicts reachability-bound, time-consuming, coverage-limited, and sensitive to the agent's gameplay ability. We present GameGen-Verifier, an automated verification paradigm for LLM-generated games that decomposes a specification into verifiable keypoints and grounds them into independent verification units. Each unit patches the game runtime into a concrete target state, executes a bounded interaction, and judges the outcome against the keypoint assertion. We implement GGV-Harness, a scalable agentic harness providing concurrency management, runtime isolation, and fault recovery. On VeriGame, our dataset of 100 games across seven genres, GameGen-Verifier achieves up to 92.2% accuracy against human judgments versus 58.8% for the coverage-enforced Agent-as-a-Verifier baseline, while reducing wall-clock time by up to 16.6x.

WeiFu Fu, CongChong Nie, Ting Sun et al.

This report introduces the technical details of the team FuXi-Fresher for LVIS Challenge 2021. Our method focuses on the problem in following two aspects: the long-tail distribution and the segmentation quality of mask and boundary. Based on the advanced HTC instance segmentation algorithm, we connect transformer backbone(Swin-L) through composite connections inspired by CBNetv2 to enhance the baseline results. To alleviate the problem of long-tail distribution, we design a Distribution Balanced method which includes dataset balanced and loss function balaced modules. Further, we use a Mask and Boundary Refinement method composed with mask scoring and refine-mask algorithms to improve the segmentation quality. In addition, we are pleasantly surprised to find that early stopping combined with EMA method can achieve a great improvement. Finally, by using multi-scale testing and increasing the upper limit of the number of objects detected per image, we achieved more than 45.4% boundary AP on the val set of LVIS Challenge 2021. On the test data of LVIS Challenge 2021, we rank 1st and achieve 48.1% AP. Notably, our APr 47.5% is very closed to the APf 48.0%.

Ting Sun, Jinlin Chen, Francis Ng

Object detection is an essential technique for autonomous driving. The performance of an object detector significantly degrades if the weather of the training images is different from that of test images. Domain adaptation can be used to address the domain shift problem so as to improve the robustness of an object detector. However, most existing domain adaptation methods either handle single target domain or require domain labels. We propose a novel unsupervised domain classification method which can be used to generalize single-target domain adaptation methods to multi-target domains, and design a weather-invariant object detector training framework based on it. We conduct the experiments on Cityscapes dataset and its synthetic variants, i.e. foggy, rainy, and night. The experimental results show that the object detector trained by our proposed method realizes robust object detection under different weather conditions.

Yu Wang, Yining Sun, Zuchang Ma et al.

Named Entity Recognition (NER) is a fundamental Natural Language Processing (NLP) task to extract entities from unstructured data. The previous methods for NER were based on machine learning or deep learning. Recently, pre-training models have significantly improved performance on multiple NLP tasks. In this paper, firstly, we introduce the architecture and pre-training tasks of four common pre-training models: BERT, ERNIE, ERNIE2.0-tiny, and RoBERTa. Then, we apply these pre-training models to a NER task by fine-tuning, and compare the effects of the different model architecture and pre-training tasks on the NER task. The experiment results showed that RoBERTa achieved state-of-the-art results on the MSRA-2006 dataset.

Ting Sun, Yuxiang Sun, Ming Liu et al.

Moving objects can greatly jeopardize the performance of a visual simultaneous localization and mapping (vSLAM) system which relies on the static-world assumption. Motion removal have seen successful on solving this problem. Two main streams of solutions are based on either geometry constraints or deep semantic segmentation neural network. The former rely on static majority assumption, and the latter require labor-intensive pixel-wise annotations. In this paper we propose to adopt a novel weakly-supervised semantic segmentation method. The segmentation mask is obtained from a CNN pre-trained with image-level class labels only. Thus, we leverage the power of deep semantic segmentation CNNs, while avoid requiring expensive annotations for training. We integrate our motion removal approach with the ORB-SLAM2 system. Experimental results on the TUM RGB-D and the KITTI stereo datasets demonstrate our superiority over the state-of-the-art.

Ting Sun, Lei Tai, Zhihan Gao et al.

This paper proposes a novel weakly-supervised semantic segmentation method using image-level label only. The class-specific activation maps from the well-trained classifiers are used as cues to train a segmentation network. The well-known defects of these cues are coarseness and incompleteness. We use super-pixel to refine them, and fuse the cues extracted from both a color image trained classifier and a gray image trained classifier to compensate for their incompleteness. The conditional random field is adapted to regulate the training process and to refine the outputs further. Besides initializing the segmentation network, the previously trained classifier is also used in the testing phase to suppress the non-existing classes. Experimental results on the PASCAL VOC 2012 dataset illustrate the effectiveness of our method.

Lianping Yang, Bin Shao, Ting Sun et al.

Most of the face hallucination methods are designed for complete inputs. They will not work well if the inputs are very tiny or contaminated by large occlusion. Inspired by this fact, we propose an obscured face hallucination network(OFHNet). The OFHNet consists of four parts: an inpainting network, an upsampling network, a discriminative network, and a fixed facial landmark detection network. The inpainting network restores the low-resolution(LR) obscured face images. The following upsampling network is to upsample the output of inpainting network. In order to ensure the generated high-resolution(HR) face images more photo-realistic, we utilize the discriminative network and the facial landmark detection network to better the result of upsampling network. In addition, we present a semantic structure loss, which makes the generated HR face images more pleasing. Extensive experiments show that our framework can restore the appealing HR face images from 1/4 missing area LR face images with a challenging scaling factor of 8x.

Ting Sun, Dezhen Song, Dit-Yan Yeung et al.

This paper presents a novel method to reduce the scale drift for indoor monocular simultaneous localization and mapping (SLAM). We leverage the prior knowledge that in the indoor environment, the line segments form tight clusters, e.g. many door frames in a straight corridor are of the same shape, size and orientation, so the same edges of these door frames form a tight line segment cluster. We implement our method in the popular ORB-SLAM2, which also serves as our baseline. In the front end we detect the line segments in each frame and incrementally cluster them in the 3D space. In the back end, we optimize the map imposing the constraint that the line segments of the same cluster should be the same. Experimental results show that our proposed method successfully reduces the scale drift for indoor monocular SLAM.

Ting Sun, Ming Liu, Haoyang Ye et al.

This paper proposes a novel point-cloud-based place recognition system that adopts a deep learning approach for feature extraction. By using a convolutional neural network pre-trained on color images to extract features from a range image without fine-tuning on extra range images, significant improvement has been observed when compared to using hand-crafted features. The resulting system is illumination invariant, rotation invariant and robust against moving objects that are unrelated to the place identity. Apart from the system itself, we also bring to the community a new place recognition dataset containing both point cloud and grayscale images covering a full $360^\circ$ environmental view. In addition, the dataset is organized in such a way that it facilitates experimental validation with respect to rotation invariance or robustness against unrelated moving objects separately.

Ting Sun, Shengyi Nie, Dit-Yan Yeung et al.

Aerial robots are becoming popular among general public, and with the development of artificial intelligence (AI), there is a trend to equip aerial robots with a natural user interface (NUI). Hand/arm gestures are an intuitive way to communicate for humans, and various research works have focused on controlling an aerial robot with natural gestures. However, the techniques in this area are still far from mature. Many issues in this area have been poorly addressed, such as the principles of choosing gestures from the design point of view, hardware requirements from an economic point of view, considerations of data availability, and algorithm complexity from a practical perspective. Our work focuses on building an economical monocular system particularly designed for gesture-based piloting of an aerial robot. Natural arm gestures are mapped to rich target directions and convenient fine adjustment is achieved. Practical piloting scenarios, hardware cost and algorithm applicability are jointly considered in our system design. The entire system is successfully implemented in an aerial robot and various properties of the system are tested.

Ting Sun, Lin Sun, Dit-Yan Yeung

Fine-grained categorization can benefit from part-based features which reveal subtle visual differences between object categories. Handcrafted features have been widely used for part detection and classification. Although a recent trend seeks to learn such features automatically using powerful deep learning models such as convolutional neural networks (CNN), their training and possibly also testing require manually provided annotations which are costly to obtain. To relax these requirements, we assume in this study a general problem setting in which the raw images are only provided with object-level class labels for model training with no other side information needed. Specifically, by extracting and interpreting the hierarchical hidden layer features learned by a CNN, we propose an elaborate CNN-based system for fine-grained categorization. When evaluated on the Caltech-UCSD Birds-200-2011, FGVC-Aircraft, Cars and Stanford dogs datasets under the setting that only object-level class labels are used for training and no other annotations are available for both training and testing, our method achieves impressive performance that is superior or comparable to the state of the art. Moreover, it sheds some light on ingenious use of the hierarchical features learned by CNN which has wide applicability well beyond the current fine-grained categorization task.