The rapid advancement of talking-head deepfake generation fueled by advanced generative models has elevated the realism of synthetic videos to a level that poses substantial risks in domains such as media, politics, and finance. However, current benchmarks for deepfake talking-head detection fail to reflect this progress, relying on outdated generators and offering limited insight into model robustness and generalization. We introduce TalkingHeadBench, a comprehensive multi-model multi-generator benchmark and curated dataset designed to evaluate the performance of state-of-the-art detectors on the most advanced generators. Our dataset includes deepfakes synthesized by leading academic and commercial models and features carefully constructed protocols to assess generalization under distribution shifts in identity and generator characteristics. We benchmark a diverse set of existing detection methods, including CNNs, vision transformers, and temporal models, and analyze their robustness and generalization capabilities. In addition, we provide error analysis using Grad-CAM visualizations to expose common failure modes and detector biases. TalkingHeadBench is hosted on https://huggingface.co/datasets/luchaoqi/TalkingHeadBench with open access to all data splits and protocols. Our benchmark aims to accelerate research towards more robust and generalizable detection models in the face of rapidly evolving generative techniques.
Head detection and tracking are essential for downstream tasks, but current methods often require large computational budgets, which increase latencies and ties up resources (e.g., processors, memory, and bandwidth). To address this, we propose a framework to enhance tiny head detection and tracking by optimizing the balance between performance and efficiency. Our framework integrates (1) a cross-domain detection loss, (2) a multi-scale module, and (3) a small receptive field detection mechanism. These innovations enhance detection by bridging the gap between large and small detectors, capturing high-frequency details at multiple scales during training, and using filters with small receptive fields to detect tiny heads. Evaluations on the CroHD and CrowdHuman datasets show improved Multiple Object Tracking Accuracy (MOTA) and mean Average Precision (mAP), demonstrating the effectiveness of our approach in crowded scenes.
Wheat plays a critical role in global food security, making it one of the most extensively studied crops. Accurate identification and measurement of key characteristics of wheat heads are essential for breeders to select varieties for cross-breeding, with the goal of developing nutrient-dense, resilient, and sustainable cultivars. Traditionally, these measurements are performed manually, which is both time-consuming and inefficient. Advances in digital technologies have paved the way for automating this process. However, field conditions pose significant challenges, such as occlusions of leaves, overlapping wheat heads, varying lighting conditions, and motion blur. In this paper, we propose a novel data augmentation technique, BBoxCut, which uses random localized masking to simulate occlusions caused by leaves and neighboring wheat heads. We evaluated our approach using three state-of-the-art object detectors and observed mean average precision (mAP) gains of 2.76, 3.26, and 1.9 for Faster R-CNN, FCOS, and DETR, respectively. Our augmentation technique led to significant improvements both qualitatively and quantitatively. In particular, the improvements were particularly evident in scenarios involving occluded wheat heads, demonstrating the robustness of our method in challenging field conditions.
The automatic detection of pedestrian heads in crowded environments is essential for crowd analysis and management tasks, particularly in high-risk settings such as railway platforms and event entrances. These environments, characterized by dense crowds and dynamic movements, are underrepresented in public datasets, posing challenges for existing deep learning models. To address this gap, we introduce the Railway Platforms and Event Entrances-Heads (RPEE-Heads) dataset, a novel, diverse, high-resolution, and accurately annotated resource. It includes 109,913 annotated pedestrian heads across 1,886 images from 66 video recordings, with an average of 56.2 heads per image. Annotations include bounding boxes for visible head regions. In addition to introducing the RPEE-Heads dataset, this paper evaluates eight state-of-the-art object detection algorithms using the RPEE-Heads dataset and analyzes the impact of head size on detection accuracy. The experimental results show that You Only Look Once v9 and Real-Time Detection Transformer outperform the other algorithms, achieving mean average precisions of 90.7% and 90.8%, with inference times of 11 and 14 milliseconds, respectively. Moreover, the findings underscore the need for specialized datasets like RPEE-Heads for training and evaluating accurate models for head detection in railway platforms and event entrances. The dataset and pretrained models are available at https://doi.org/10.34735/ped.2024.2.
Pedestrian detection and tracking in crowded video sequences have a wide range of applications, including autonomous driving, robot navigation and pedestrian flow surveillance. However, detecting and tracking pedestrians in high-density crowds face many challenges, including intra-class occlusions, complex motions, and diverse poses. Although deep learning models have achieved remarkable progress in head detection, head tracking datasets and methods are extremely lacking. Existing head datasets have limited coverage of complex pedestrian flows and scenes (e.g., pedestrian interactions, occlusions, and object interference). It is of great importance to develop new head tracking datasets and methods. To address these challenges, we present a Chinese Large-scale Cross-scene Pedestrian Head Tracking dataset (Cchead) and a Multi-Source Information Fusion Network (MIFN). Our dataset has features that are of considerable interest, including 10 diverse scenes of 50,528 frames with over 2,366,249 heads and 2,358 tracks annotated. Our dataset contains diverse human moving speeds, directions, and complex crowd pedestrian flows with collision avoidance behaviors. We provide a comprehensive analysis and comparison with existing state-of-the-art (SOTA) algorithms. Moreover, our MIFN is the first end-to-end CNN-based head detection and tracking network that jointly trains RGB frames, pixel-level motion information (optical flow and frame difference maps), depth maps, and density maps in videos. Compared with SOTA pedestrian detection and tracking methods, MIFN achieves superior performance on our Cchead dataset. We believe our datasets and baseline will become valuable resources towards developing pedestrian tracking in dense crowds.
Human head detection, keypoint estimation, and 3D head model fitting are essential tasks with many applications. However, traditional real-world datasets often suffer from bias, privacy, and ethical concerns, and they have been recorded in laboratory environments, which makes it difficult for trained models to generalize. Here, we introduce \method -- a large-scale synthetic dataset generated with diffusion models for human head detection and 3D mesh estimation. Our dataset comprises over 1 million high-resolution images, each annotated with detailed 3D head meshes, facial landmarks, and bounding boxes. Using this dataset, we introduce a new model architecture capable of simultaneous head detection and head mesh reconstruction from a single image in a single step. Through extensive experimental evaluations, we demonstrate that models trained on our synthetic data achieve strong performance on real images. Furthermore, the versatility of our dataset makes it applicable across a broad spectrum of tasks, offering a general and comprehensive representation of human heads.
Deep neural networks (DNNs) have exhibited remarkable performance across various tasks, yet their susceptibility to spurious correlations poses a significant challenge for out-of-distribution (OOD) generalization. Spurious correlations refer to erroneous associations in data that do not reflect true underlying relationships but are instead artifacts of dataset characteristics or biases. These correlations can lead DNNs to learn patterns that are not robust across diverse datasets or real-world scenarios, hampering their ability to generalize beyond training data. In this paper, we propose an autoencoder-based approach to analyze the nature of spurious correlations that exist in the Global Wheat Head Detection (GWHD) 2021 dataset. We then use inpainting followed by Weighted Boxes Fusion (WBF) to achieve a 2% increase in the Average Domain Accuracy (ADA) over the YOLOv5 baseline and consistently show that our approach has the ability to suppress some of the spurious correlations in the GWHD 2021 dataset. The key advantage of our approach is that it is more suitable in scenarios where there is limited scope to adapt or fine-tune the trained model in unseen test environments.
Optic nerve head (ONH) detection has been a crucial area of study in ophthalmology for years. However, the significant discrepancy between fundus image datasets, each generated using a single type of fundus camera, poses challenges to the generalizability of ONH detection approaches developed based on semantic segmentation networks. Despite the numerous recent advancements in general-purpose semantic segmentation methods using convolutional neural networks (CNNs) and Transformers, there is currently a lack of benchmarks for these state-of-the-art (SoTA) networks specifically trained for ONH detection. Therefore, in this article, we make contributions from three key aspects: network design, the publication of a dataset, and the establishment of a comprehensive benchmark. Our newly developed ONH detection network, referred to as ODFormer, is based upon the Swin Transformer architecture and incorporates two novel components: a multi-scale context aggregator and a lightweight bidirectional feature recalibrator. Our published large-scale dataset, known as TongjiU-DROD, provides multi-resolution fundus images for each participant, captured using two distinct types of cameras. Our established benchmark involves three datasets: DRIONS-DB, DRISHTI-GS1, and TongjiU-DROD, created by researchers from different countries and containing fundus images captured from participants of diverse races and ages. Extensive experimental results demonstrate that our proposed ODFormer outperforms other state-of-the-art (SoTA) networks in terms of performance and generalizability. Our dataset and source code are publicly available at mias.group/ODFormer.
The diverse and high-quality content generated by recent generative models demonstrates the great potential of using synthetic data to train downstream models. However, in vision, especially in objection detection, related areas are not fully explored, the synthetic images are merely used to balance the long tails of existing datasets, and the accuracy of the generated labels is low, the full potential of generative models has not been exploited. In this paper, we propose DODA, a data synthesizer that can generate high-quality object detection data for new domains in agriculture. Specifically, we improve the controllability of layout-to-image through encoding layout as an image, thereby improving the quality of labels, and use a visual encoder to provide visual clues for the diffusion model to decouple visual features from the diffusion model, and empowering the model the ability to generate data in new domains. On the Global Wheat Head Detection (GWHD) Dataset, which is the largest dataset in agriculture and contains diverse domains, using the data synthesized by DODA improves the performance of the object detector by 12.74-17.76 AP$_{50}$ in the domain that was significantly shifted from the training data.
Deep learning models have been used for a variety of image processing tasks. However, most of these models are developed through supervised learning approaches, which rely heavily on the availability of large-scale annotated datasets. Developing such datasets is tedious and expensive. In the absence of an annotated dataset, synthetic data can be used for model development; however, due to the substantial differences between simulated and real data, a phenomenon referred to as domain gap, the resulting models often underperform when applied to real data. In this research, we aim to address this challenge by first computationally simulating a large-scale annotated dataset and then using a generative adversarial network (GAN) to fill the gap between simulated and real images. This approach results in a synthetic dataset that can be effectively utilized to train a deep-learning model. Using this approach, we developed a realistic annotated synthetic dataset for wheat head segmentation. This dataset was then used to develop a deep-learning model for semantic segmentation. The resulting model achieved a Dice score of 83.4\% on an internal dataset and Dice scores of 79.6% and 83.6% on two external Global Wheat Head Detection datasets. While we proposed this approach in the context of wheat head segmentation, it can be generalized to other crop types or, more broadly, to images with dense, repeated patterns such as those found in cellular imagery.
Head detection provides distribution information of pedestrian, which is crucial for scene statistical analysis, traffic management, and risk assessment and early warning. However, scene complexity and large-scale variation in the real world make accurate detection more difficult. Therefore, we present a modified Yolov8 which improves head detection performance through reinforcing target perception. An Auxiliary Learning Feature Fusion (ALFF) module comprised of LSTM and convolutional blocks is used as the auxiliary task to help the model perceive targets. In addition, we introduce Noise Calibration into Distribution Focal Loss to facilitate model fitting and improve the accuracy of detection. Considering the requirements of high accuracy and speed for the head detection task, our method is adapted with two kinds of backbone, namely Yolov8n and Yolov8m. The results demonstrate the superior performance of our approach in improving detection accuracy and robustness.
Detection of human body and its parts has been intensively studied. However, most of CNNs-based detectors are trained independently, making it difficult to associate detected parts with body. In this paper, we focus on the joint detection of human body and its parts. Specifically, we propose a novel extended object representation integrating center-offsets of body parts, and construct an end-to-end generic Body-Part Joint Detector (BPJDet). In this way, body-part associations are neatly embedded in a unified representation containing both semantic and geometric contents. Therefore, we can optimize multi-loss to tackle multi-tasks synergistically. Moreover, this representation is suitable for anchor-based and anchor-free detectors. BPJDet does not suffer from error-prone post matching, and keeps a better trade-off between speed and accuracy. Furthermore, BPJDet can be generalized to detect body-part or body-parts of either human or quadruped animals. To verify the superiority of BPJDet, we conduct experiments on datasets of body-part (CityPersons, CrowdHuman and BodyHands) and body-parts (COCOHumanParts and Animals5C). While keeping high detection accuracy, BPJDet achieves state-of-the-art association performance on all datasets. Besides, we show benefits of advanced body-part association capability by improving performance of two representative downstream applications: accurate crowd head detection and hand contact estimation. Project is available in https://hnuzhy.github.io/projects/BPJDet.
Wheat is one of the most significant crop species with an annual worldwide grain production of 700 million tonnes. Assessing the production of wheat spikes can help us measure the grain production. Thus, detecting and characterizing spikes from images of wheat fields is an essential component in a wheat breeding process. In this study, we propose three wheat head counting networks (WHCNet\_1, WHCNet\_2 and WHCNet\_3) to accurately estimate the wheat head count from an individual image and construct high quality density map, which illustrates the distribution of wheat heads in the image. The WHCNets are composed of two major components: a convolutional neural network (CNN) as the front-end for wheat head image feature extraction and a CNN with skip connections for the back-end to generate high-quality density maps. The dataset used in this study is the Global Wheat Head Detection (GWHD) dataset, which is a large, diverse, and well-labelled dataset of wheat images and built by a joint international collaborative effort. We compare our methods with CSRNet, a deep learning method which developed for highly congested scenes understanding and performing accurate count estimation as well as presenting high quality density maps. By taking the advantage of the skip connections between CNN layers, WHCNets integrate features from low CNN layers to high CNN layers, thus, the output density maps have both high spatial resolution and detailed representations of the input images. The experiments showed that our methods outperformed CSRNet in terms of the evaluation metrics, mean absolute error (MAE) and the root mean squared error (RMSE) with smaller model sizes. The code has been deposited on GitHub (\url{https://github.com/hyguozz}).
Existing head pose estimation (HPE) mainly focuses on single person with pre-detected frontal heads, which limits their applications in real complex scenarios with multi-persons. We argue that these single HPE methods are fragile and inefficient for Multi-Person Head Pose Estimation (MPHPE) since they rely on the separately trained face detector that cannot generalize well to full viewpoints, especially for heads with invisible face areas. In this paper, we focus on the full-range MPHPE problem, and propose a direct end-to-end simple baseline named DirectMHP. Due to the lack of datasets applicable to the full-range MPHPE, we firstly construct two benchmarks by extracting ground-truth labels for head detection and head orientation from public datasets AGORA and CMU Panoptic. They are rather challenging for having many truncated, occluded, tiny and unevenly illuminated human heads. Then, we design a novel end-to-end trainable one-stage network architecture by joint regressing locations and orientations of multi-head to address the MPHPE problem. Specifically, we regard pose as an auxiliary attribute of the head, and append it after the traditional object prediction. Arbitrary pose representation such as Euler angles is acceptable by this flexible design. Then, we jointly optimize these two tasks by sharing features and utilizing appropriate multiple losses. In this way, our method can implicitly benefit from more surroundings to improve HPE accuracy while maintaining head detection performance. We present comprehensive comparisons with state-of-the-art single HPE methods on public benchmarks, as well as superior baseline results on our constructed MPHPE datasets. Datasets and code are released in https://github.com/hnuzhy/DirectMHP.
As a fundamental computer vision task, crowd counting plays an important role in public safety. Currently, deep learning based head detection is a promising method for crowd counting. However, the highly concerned object detection networks cannot be well applied to this problem for three reasons: (1) Existing loss functions fail to address sample imbalance in highly dense and complex scenes; (2) Canonical object detectors lack spatial coherence in loss calculation, disregarding the relationship between object location and background region; (3) Most of the head detection datasets are only annotated with the center points, i.e. without bounding boxes. To overcome these issues, we propose a novel Mask Focal Loss (MFL) based on heatmap via the Gaussian kernel. MFL provides a unifying framework for the loss functions based on both heatmap and binary feature map ground truths. Additionally, we introduce GTA_Head, a synthetic dataset with comprehensive annotations, for evaluation and comparison. Extensive experimental results demonstrate the superior performance of our MFL across various detectors and datasets, and it can reduce MAE and RMSE by up to 47.03% and 61.99%, respectively. Therefore, our work presents a strong foundation for advancing crowd counting methods based on density estimation.
Most recent head pose estimation (HPE) methods are dominated by the Euler angle representation. To avoid its inherent ambiguity problem of rotation labels, alternative quaternion-based and vector-based representations are introduced. However, they both are not visually intuitive, and often derived from equivocal Euler angle labels. In this paper, we present a novel single-stage keypoint-based method via an {\it intuitive} and {\it unconstrained} 2D cube representation for joint head detection and pose estimation. The 2D cube is an orthogonal projection of the 3D regular hexahedron label roughly surrounding one head, and itself contains the head location. It can reflect the head orientation straightforwardly and unambiguously in any rotation angle. Unlike the general 6-DoF object pose estimation, our 2D cube ignores the 3-DoF of head size but retains the 3-DoF of head pose. Based on the prior of equal side length, we can effortlessly obtain the closed-form solution of Euler angles from predicted 2D head cube instead of applying the error-prone PnP algorithm. In experiments, our proposed method achieves comparable results with other representative methods on the public AFLW2000 and BIWI datasets. Besides, a novel test on the CMU panoptic dataset shows that our method can be seamlessly adapted to the unconstrained full-view HPE task without modification.
The automatic detection and localization of anatomical features in retinal imaging data are relevant for many aspects. In this work, we follow a data-centric approach to optimize classifier training for optic nerve head detection and localization in optical coherence tomography en face images of the retina. We examine the effect of domain knowledge driven spatial complexity reduction on the resulting optic nerve head segmentation and localization performance. We present a machine learning approach for segmenting optic nerve head in 2D en face projections of 3D widefield swept source optical coherence tomography scans that enables the automated assessment of large amounts of data. Evaluation on manually annotated 2D en face images of the retina demonstrates that training of a standard U-Net can yield improved optic nerve head segmentation and localization performance when the underlying pixel-level binary classification task is spatially relaxed through domain knowledge.
In this paper, we study the cluster head detection problem of a two-level unmanned aerial vehicle (UAV) swarm network (USNET) with multiple UAV clusters, where the inherent follow strategy (IFS) of low-level follower UAVs (FUAVs) with respect to high-level cluster head UAVs (HUAVs) is unknown. We first propose a graph attention self-supervised learning algorithm (GASSL) to detect the HUAVs of a single UAV cluster, where the GASSL can fit the IFS at the same time. Then, to detect the HUAVs in the USNET with multiple UAV clusters, we develop a multi-cluster graph attention self-supervised learning algorithm (MC-GASSL) based on the GASSL. The MC-GASSL clusters the USNET with a gated recurrent unit (GRU)-based metric learning scheme and finds the HUAVs in each cluster with GASSL. Numerical results show that the GASSL can detect the HUAVs in single UAV clusters obeying various kinds of IFSs with over 98% average accuracy. The simulation results also show that the clustering purity of the USNET with MC-GASSL exceeds that with traditional clustering algorithms by at least 10% average. Furthermore, the MC-GASSL can efficiently detect all the HUAVs in USNETs with various IFSs and cluster numbers with low detection redundancies.
The mortality of lung cancer has ranked high among cancers for many years. Early detection of lung cancer is critical for disease prevention, cure, and mortality rate reduction. However, existing detection methods on pulmonary nodules introduce an excessive number of false positive proposals in order to achieve high sensitivity, which is not practical in clinical situations. In this paper, we propose the multi-head detection and spatial squeeze-and-attention network, MHSnet, to detect pulmonary nodules, in order to aid doctors in the early diagnosis of lung cancers. Specifically, we first introduce multi-head detectors and skip connections to customize for the variety of nodules in sizes, shapes and types and capture multi-scale features. Then, we implement a spatial attention module to enable the network to focus on different regions differently inspired by how experienced clinicians screen CT images, which results in fewer false positive proposals. Lastly, we present a lightweight but effective false positive reduction module with the Linear Regression model to cut down the number of false positive proposals, without any constraints on the front network. Extensive experimental results compared with the state-of-the-art models have shown the superiority of the MHSnet in terms of the average FROC, sensitivity and especially false discovery rate (2.98% and 2.18% improvement in terms of average FROC and sensitivity, 5.62% and 28.33% decrease in terms of false discovery rate and average candidates per scan). The false positive reduction module significantly decreases the average number of candidates generated per scan by 68.11% and the false discovery rate by 13.48%, which is promising to reduce distracted proposals for the downstream tasks based on the detection results.
We study the problem of detecting talking activities in collaborative learning videos. Our approach uses head detection and projections of the log-magnitude of optical flow vectors to reduce the problem to a simple classification of small projection images without the need for training complex, 3-D activity classification systems. The small projection images are then easily classified using a simple majority vote of standard classifiers. For talking detection, our proposed approach is shown to significantly outperform single activity systems. We have an overall accuracy of 59% compared to 42% for Temporal Segment Network (TSN) and 45% for Convolutional 3D (C3D). In addition, our method is able to detect multiple talking instances from multiple speakers, while also detecting the speakers themselves.