Search

Existing generic unsupervised domain adaptation approaches require access to both a large labeled source dataset and a sufficient unlabeled target dataset during adaptation. However, collecting a large dataset, even if unlabeled, is a challenging and expensive endeavor, especially in medical imaging. In addition, constraints such as privacy issues can result in cases where source data is unavailable. Taking in consideration these challenges, we propose MIAdapt, an adaptive approach for Microscopic Imagery Adaptation as a solution for Source-free Few-shot Domain Adaptive Object detection (SF-FSDA). We also define two competitive baselines (1) Faster-FreeShot and (2) MT-FreeShot. Extensive experiments on the challenging M5-Malaria and Raabin-WBC datasets validate the effectiveness of MIAdapt. Without using any image from the source domain MIAdapt surpasses state-of-the-art source-free UDA (SF-UDA) methods by +21.3% mAP and few-shot domain adaptation (FSDA) approaches by +4.7% mAP on Raabin-WBC. Our code and models will be publicly available., Comment: 6 pages, 5 figures

Compositional Zero-Shot Learning (CZSL) aims to predict unknown compositions made up of attribute and object pairs. Predicting compositions unseen during training is a challenging task. We are exploring Open World Compositional Zero-Shot Learning (OW-CZSL) in this study, where our test space encompasses all potential combinations of attributes and objects. Our approach involves utilizing the self-attention mechanism between attributes and objects to achieve better generalization from seen to unseen compositions. Utilizing a self-attention mechanism facilitates the model's ability to identify relationships between attribute and objects. The similarity between the self-attended textual and visual features is subsequently calculated to generate predictions during the inference phase. The potential test space may encompass implausible object-attribute combinations arising from unrestricted attribute-object pairings. To mitigate this issue, we leverage external knowledge from ConceptNet to restrict the test space to realistic compositions. Our proposed model, Attention-based Simple Primitives (ASP), demonstrates competitive performance, achieving results comparable to the state-of-the-art., Comment: 10 pages, 6 figures

In recent years, numerous domain adaptive strategies have been proposed to help deep learning models overcome the challenges posed by domain shift. However, even unsupervised domain adaptive strategies still require a large amount of target data. Medical imaging datasets are often characterized by class imbalance and scarcity of labeled and unlabeled data. Few-shot domain adaptive object detection (FSDAOD) addresses the challenge of adapting object detectors to target domains with limited labeled data. Existing works struggle with randomly selected target domain images that may not accurately represent the real population, resulting in overfitting to small validation sets and poor generalization to larger test sets. Medical datasets exhibit high class imbalance and background similarity, leading to increased false positives and lower mean Average Precision (map) in target domains. To overcome these challenges, we propose a novel FSDAOD strategy for microscopic imaging. Our contributions include a domain adaptive class balancing strategy for few-shot scenarios, multi-layer instance-level inter and intra-domain alignment to enhance similarity between class instances regardless of domain, and an instance-level classification loss applied in the middle layers of the object detector to enforce feature retention necessary for correct classification across domains. Extensive experimental results with competitive baselines demonstrate the effectiveness of our approach, achieving state-of-the-art results on two public microscopic datasets. Code available at https://github.co/intelligentMachinesLab/few-shot-domain-adaptive-microscopy, Comment: Accepted to MICCAI 2024 main conference

In this work, we tackle the problem of domain generalization for object detection, specifically focusing on the scenario where only a single source domain is available. We propose an effective approach that involves two key steps: diversifying the source domain and aligning detections based on class prediction confidence and localization. Firstly, we demonstrate that by carefully selecting a set of augmentations, a base detector can outperform existing methods for single domain generalization by a good margin. This highlights the importance of domain diversification in improving the performance of object detectors. Secondly, we introduce a method to align detections from multiple views, considering both classification and localization outputs. This alignment procedure leads to better generalized and well-calibrated object detector models, which are crucial for accurate decision-making in safety-critical applications. Our approach is detector-agnostic and can be seamlessly applied to both single-stage and two-stage detectors. To validate the effectiveness of our proposed methods, we conduct extensive experiments and ablations on challenging domain-shift scenarios. The results consistently demonstrate the superiority of our approach compared to existing methods. Our code and models are available at: https://github.com/msohaildanish/DivAlign

Earlier diagnosis of Leukemia can save thousands of lives annually. The prognosis of leukemia is challenging without the morphological information of White Blood Cells (WBC) and relies on the accessibility of expensive microscopes and the availability of hematologists to analyze Peripheral Blood Samples (PBS). Deep Learning based methods can be employed to assist hematologists. However, these algorithms require a large amount of labeled data, which is not readily available. To overcome this limitation, we have acquired a realistic, generalized, and large dataset. To collect this comprehensive dataset for real-world applications, two microscopes from two different cost spectrums (high-cost HCM and low-cost LCM) are used for dataset capturing at three magnifications (100x, 40x, 10x) through different sensors (high-end camera for HCM, middle-level camera for LCM and mobile-phone camera for both). The high-sensor camera is 47 times more expensive than the middle-level camera and HCM is 17 times more expensive than LCM. In this collection, using HCM at high resolution (100x), experienced hematologists annotated 10.3k WBC types (14) and artifacts, having 55k morphological labels (Cell Size, Nuclear Chromatin, Nuclear Shape, etc.) from 2.4k images of several PBS leukemia patients. Later on, these annotations are transferred to other 2 magnifications of HCM, and 3 magnifications of LCM, and on each camera captured images. Along with the LeukemiaAttri dataset, we provide baselines over multiple object detectors and Unsupervised Domain Adaptation (UDA) strategies, along with morphological information-based attribute prediction. The dataset will be publicly available after publication to facilitate the research in this direction., Comment: Early Accept

Deep learning based object detectors struggle generalizing to a new target domain bearing significant variations in object and background. Most current methods align domains by using image or instance-level adversarial feature alignment. This often suffers due to unwanted background and lacks class-specific alignment. A straightforward approach to promote class-level alignment is to use high confidence predictions on unlabeled domain as pseudo-labels. These predictions are often noisy since model is poorly calibrated under domain shift. In this paper, we propose to leverage model's predictive uncertainty to strike the right balance between adversarial feature alignment and class-level alignment. We develop a technique to quantify predictive uncertainty on class assignments and bounding-box predictions. Model predictions with low uncertainty are used to generate pseudo-labels for self-training, whereas the ones with higher uncertainty are used to generate tiles for adversarial feature alignment. This synergy between tiling around uncertain object regions and generating pseudo-labels from highly certain object regions allows capturing both image and instance-level context during the model adaptation. We report thorough ablation study to reveal the impact of different components in our approach. Results on five diverse and challenging adaptation scenarios show that our approach outperforms existing state-of-the-art methods with noticeable margins., Comment: Accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (Volume: 45, Issue: 12, December 2023); Extended version of our conference paper, arXiv link: arXiv:2110.00249

Albeit revealing impressive predictive performance for several computer vision tasks, deep neural networks (DNNs) are prone to making overconfident predictions. This limits the adoption and wider utilization of DNNs in many safety-critical applications. There have been recent efforts toward calibrating DNNs, however, almost all of them focus on the classification task. Surprisingly, very little attention has been devoted to calibrating modern DNN-based object detectors, especially detection transformers, which have recently demonstrated promising detection performance and are influential in many decision-making systems. In this work, we address the problem by proposing a mechanism for calibrated detection transformers (Cal-DETR), particularly for Deformable-DETR, UP-DETR and DINO. We pursue the train-time calibration route and make the following contributions. First, we propose a simple yet effective approach for quantifying uncertainty in transformer-based object detectors. Second, we develop an uncertainty-guided logit modulation mechanism that leverages the uncertainty to modulate the class logits. Third, we develop a logit mixing approach that acts as a regularizer with detection-specific losses and is also complementary to the uncertainty-guided logit modulation technique to further improve the calibration performance. Lastly, we conduct extensive experiments across three in-domain and four out-domain scenarios. Results corroborate the effectiveness of Cal-DETR against the competing train-time methods in calibrating both in-domain and out-domain detections while maintaining or even improving the detection performance. Our codebase and pre-trained models can be accessed at \url{https://github.com/akhtarvision/cal-detr}., Comment: Accepted at NeurIPS 2023

Getting precise aspects of road through segmentation from remote sensing imagery is useful for many real-world applications such as autonomous vehicles, urban development and planning, and achieving sustainable development goals. Roads are only a small part of the image, and their appearance, type, width, elevation, directions, etc. exhibit large variations across geographical areas. Furthermore, due to differences in urbanization styles, planning, and the natural environments; regions along the roads vary significantly. Due to these variations among the train and test domains, the road segmentation algorithms fail to generalize to new geographical locations. Unlike the generic domain alignment scenarios, road segmentation has no scene structure, and generic domain adaptation methods are unable to enforce topological properties like continuity, connectivity, smoothness, etc., thus resulting in degraded domain alignment. In this work, we propose a topology-aware unsupervised domain adaptation approach for road segmentation in remote sensing imagery. Specifically, we predict road skeleton, an auxiliary task to impose the topological constraints. To enforce consistent predictions of road and skeleton, especially in the unlabeled target domain, the conformity loss is defined across the skeleton prediction head and the road-segmentation head. Furthermore, for self-training, we filter out the noisy pseudo-labels by using a connectivity-based pseudo-labels refinement strategy, on both road and skeleton segmentation heads, thus avoiding holes and discontinuities. Extensive experiments on the benchmark datasets show the effectiveness of the proposed approach compared to existing state-of-the-art methods. Specifically, for SpaceNet to DeepGlobe adaptation, the proposed approach outperforms the competing methods by a minimum margin of 6.6%, 6.7%, and 9.8% in IoU, F1-score, and APLS, respectively.

Detecting firearms and accurately localizing individuals carrying them in images or videos is of paramount importance in security, surveillance, and content customization. However, this task presents significant challenges in complex environments due to clutter and the diverse shapes of firearms. To address this problem, we propose a novel approach that leverages human-firearm interaction information, which provides valuable clues for localizing firearm carriers. Our approach incorporates an attention mechanism that effectively distinguishes humans and firearms from the background by focusing on relevant areas. Additionally, we introduce a saliency-driven locality-preserving constraint to learn essential features while preserving foreground information in the input image. By combining these components, our approach achieves exceptional results on a newly proposed dataset. To handle inputs of varying sizes, we pass paired human-firearm instances with attention masks as channels through a deep network for feature computation, utilizing an adaptive average pooling layer. We extensively evaluate our approach against existing methods in human-object interaction detection and achieve significant results (AP=77.8\%) compared to the baseline approach (AP=63.1\%). This demonstrates the effectiveness of leveraging attention mechanisms and saliency-driven locality preservation for accurate human-firearm interaction detection. Our findings contribute to advancing the fields of security and surveillance, enabling more efficient firearm localization and identification in diverse scenarios., Comment: This paper is accepted in IEEE Transactions on Computational Social Systems

With deep neural network based solution more readily being incorporated in real-world applications, it has been pressing requirement that predictions by such models, especially in safety-critical environments, be highly accurate and well-calibrated. Although some techniques addressing DNN calibration have been proposed, they are only limited to visual classification applications and in-domain predictions. Unfortunately, very little to no attention is paid towards addressing calibration of DNN-based visual object detectors, that occupy similar space and importance in many decision making systems as their visual classification counterparts. In this work, we study the calibration of DNN-based object detection models, particularly under domain shift. To this end, we first propose a new, plug-and-play, train-time calibration loss for object detection (coined as TCD). It can be used with various application-specific loss functions as an auxiliary loss function to improve detection calibration. Second, we devise a new implicit technique for improving calibration in self-training based domain adaptive detectors, featuring a new uncertainty quantification mechanism for object detection. We demonstrate TCD is capable of enhancing calibration with notable margins (1) across different DNN-based object detection paradigms both in in-domain and out-of-domain predictions, and (2) in different domain-adaptive detectors across challenging adaptation scenarios. Finally, we empirically show that our implicit calibration technique can be used in tandem with TCD during adaptation to further boost calibration in diverse domain shift scenarios., Comment: To appear in NeurIPS 2022

This paper proposes a novel pixel-level distribution regularization scheme (DRSL) for self-supervised domain adaptation of semantic segmentation. In a typical setting, the classification loss forces the semantic segmentation model to greedily learn the representations that capture inter-class variations in order to determine the decision (class) boundary. Due to the domain shift, this decision boundary is unaligned in the target domain, resulting in noisy pseudo labels adversely affecting self-supervised domain adaptation. To overcome this limitation, along with capturing inter-class variation, we capture pixel-level intra-class variations through class-aware multi-modal distribution learning (MMDL). Thus, the information necessary for capturing the intra-class variations is explicitly disentangled from the information necessary for inter-class discrimination. Features captured thus are much more informative, resulting in pseudo-labels with low noise. This disentanglement allows us to perform separate alignments in discriminative space and multi-modal distribution space, using cross-entropy based self-learning for the former. For later, we propose a novel stochastic mode alignment method, by explicitly decreasing the distance between the target and source pixels that map to the same mode. The distance metric learning loss, computed over pseudo-labels and backpropagated from multi-modal modeling head, acts as the regularizer over the base network shared with the segmentation head. The results from comprehensive experiments on synthetic to real domain adaptation setups, i.e., GTA-V/SYNTHIA to Cityscapes, show that DRSL outperforms many existing approaches (a minimum margin of 2.3% and 2.5% in mIoU for SYNTHIA to Cityscapes)., Comment: Accepted for publication at Image and Vision Computing

One billion people worldwide are estimated to be living in slums, and documenting and analyzing these regions is a challenging task. As compared to regular slums; the small, scattered and temporary nature of temporary slums makes data collection and labeling tedious and time-consuming. To tackle this challenging problem of temporary slums detection, we present a semi-supervised deep learning segmentation-based approach; with the strategy to detect initial seed images in the zero-labeled data settings. A small set of seed samples (32 in our case) are automatically discovered by analyzing the temporal changes, which are manually labeled to train a segmentation and representation learning module. The segmentation module gathers high dimensional image representations, and the representation learning module transforms image representations into embedding vectors. After that, a scoring module uses the embedding vectors to sample images from a large pool of unlabeled images and generates pseudo-labels for the sampled images. These sampled images with their pseudo-labels are added to the training set to update the segmentation and representation learning modules iteratively. To analyze the effectiveness of our technique, we construct a large geographically marked dataset of temporary slums. This dataset constitutes more than 200 potential temporary slum locations (2.28 square kilometers) found by sieving sixty-eight thousand images from 12 metropolitan cities of Pakistan covering 8000 square kilometers. Furthermore, our proposed method outperforms several competitive semi-supervised semantic segmentation baselines on a similar setting. The code and the dataset will be made publicly available.

This paper presents FogAdapt, a novel approach for domain adaptation of semantic segmentation for dense foggy scenes. Although significant research has been directed to reduce the domain shift in semantic segmentation, adaptation to scenes with adverse weather conditions remains an open question. Large variations in the visibility of the scene due to weather conditions, such as fog, smog, and haze, exacerbate the domain shift, thus making unsupervised adaptation in such scenarios challenging. We propose a self-entropy and multi-scale information augmented self-supervised domain adaptation method (FogAdapt) to minimize the domain shift in foggy scenes segmentation. Supported by the empirical evidence that an increase in fog density results in high self-entropy for segmentation probabilities, we introduce a self-entropy based loss function to guide the adaptation method. Furthermore, inferences obtained at different image scales are combined and weighted by the uncertainty to generate scale-invariant pseudo-labels for the target domain. These scale-invariant pseudo-labels are robust to visibility and scale variations. We evaluate the proposed model on real clear-weather scenes to real foggy scenes adaptation and synthetic non-foggy images to real foggy scenes adaptation scenarios. Our experiments demonstrate that FogAdapt significantly outperforms the current state-of-the-art in semantic segmentation of foggy images. Specifically, by considering the standard settings compared to state-of-the-art (SOTA) methods, FogAdapt gains 3.8% on Foggy Zurich, 6.0% on Foggy Driving-dense, and 3.6% on Foggy Driving in mIoU when adapted from Cityscapes to Foggy Zurich., Comment: Accepted at Elsevier Journal of Neurocomputing

Malaria, a fatal but curable disease claims hundreds of thousands of lives every year. Early and correct diagnosis is vital to avoid health complexities, however, it depends upon the availability of costly microscopes and trained experts to analyze blood-smear slides. Deep learning-based methods have the potential to not only decrease the burden of experts but also improve diagnostic accuracy on low-cost microscopes. However, this is hampered by the absence of a reasonable size dataset. One of the most challenging aspects is the reluctance of the experts to annotate the dataset at low magnification on low-cost microscopes. We present a dataset to further the research on malaria microscopy over the low-cost microscopes at low magnification. Our large-scale dataset consists of images of blood-smear slides from several malaria-infected patients, collected through microscopes at two different cost spectrums and multiple magnifications. Malarial cells are annotated for the localization and life-stage classification task on the images collected through the high-cost microscope at high magnification. We design a mechanism to transfer these annotations from the high-cost microscope at high magnification to the low-cost microscope, at multiple magnifications. Multiple object detectors and domain adaptation methods are presented as the baselines. Furthermore, a partially supervised domain adaptation method is introduced to adapt the object-detector to work on the images collected from the low-cost microscope. The dataset will be made publicly available after publication.

Deep neural networks have shown promising results in disease detection and classification using medical image data. However, they still suffer from the challenges of handling real-world scenarios especially reliably detecting out-of-distribution (OoD) samples. We propose an approach to robustly classify OoD samples in skin and malaria images without the need to access labeled OoD samples during training. Specifically, we use metric learning along with logistic regression to force the deep networks to learn much rich class representative features. To guide the learning process against the OoD examples, we generate ID similar-looking examples by either removing class-specific salient regions in the image or permuting image parts and distancing them away from in-distribution samples. During inference time, the K-reciprocal nearest neighbor is employed to detect out-of-distribution samples. For skin cancer OoD detection, we employ two standard benchmark skin cancer ISIC datasets as ID, and six different datasets with varying difficulty levels were taken as out of distribution. For malaria OoD detection, we use the BBBC041 malaria dataset as ID and five different challenging datasets as out of distribution. We achieved state-of-the-art results, improving 5% and 4% in TNR@TPR95% over the previous state-of-the-art for skin cancer and malaria OoD detection respectively.

Estimating the number of buildings in any geographical region is a vital component of urban analysis, disaster management, and public policy decision. Deep learning methods for building localization and counting in satellite imagery, can serve as a viable and cheap alternative. However, these algorithms suffer performance degradation when applied to the regions on which they have not been trained. Current large datasets mostly cover the developed regions and collecting such datasets for every region is a costly, time-consuming, and difficult endeavor. In this paper, we propose an unsupervised domain adaptation method for counting buildings where we use a labeled source domain (developed regions) and adapt the trained model on an unlabeled target domain (developing regions). We initially align distribution maps across domains by aligning the output space distribution through adversarial loss. We then exploit counting consistency constraints, within-image count consistency, and across-image count consistency, to decrease the domain shift. Within-image consistency enforces that building count in the whole image should be greater than or equal to count in any of its sub-image. Across-image consistency constraint enforces that if an image contains considerably more buildings than the other image, then their sub-images shall also have the same order. These two constraints encourage the behavior to be consistent across and within the images, regardless of the scale. To evaluate the performance of our proposed approach, we collected and annotated a large-scale dataset consisting of challenging South Asian regions having higher building densities and irregular structures as compared to existing datasets. We perform extensive experiments to verify the efficacy of our approach and report improvements of approximately 7% to 20% over the competitive baseline methods.

We study adapting trained object detectors to unseen domains manifesting significant variations of object appearance, viewpoints and backgrounds. Most current methods align domains by either using image or instance-level feature alignment in an adversarial fashion. This often suffers due to the presence of unwanted background and as such lacks class-specific alignment. A common remedy to promote class-level alignment is to use high confidence predictions on the unlabelled domain as pseudo labels. These high confidence predictions are often fallacious since the model is poorly calibrated under domain shift. In this paper, we propose to leverage model predictive uncertainty to strike the right balance between adversarial feature alignment and class-level alignment. Specifically, we measure predictive uncertainty on class assignments and the bounding box predictions. Model predictions with low uncertainty are used to generate pseudo-labels for self-supervision, whereas the ones with higher uncertainty are used to generate tiles for an adversarial feature alignment stage. This synergy between tiling around the uncertain object regions and generating pseudo-labels from highly certain object regions allows us to capture both the image and instance level context during the model adaptation stage. We perform extensive experiments covering various domain shift scenarios. Our approach improves upon existing state-of-the-art methods with visible margins., Comment: To appear in NeurIPS2021

Malaria microscopy, microscopic examination of stained blood slides to detect parasite Plasmodium, is considered to be a gold-standard for detecting life-threatening disease malaria. Detecting the plasmodium parasite requires a skilled examiner and may take up to 10 to 15 minutes to completely go through the whole slide. Due to a lack of skilled medical professionals in the underdeveloped or resource deficient regions, many cases go misdiagnosed; resulting in unavoidable complications and/or undue medication. We propose to complement the medical professionals by creating a deep learning-based method to automatically detect (localize) the plasmodium parasites in the photograph of stained film. To handle the unbalanced nature of the dataset, we adopt a two-stage approach. Where the first stage is trained to detect blood cells and classify them into just healthy or infected. The second stage is trained to classify each detected cell further into the life-cycle stage. To facilitate the research in machine learning-based malaria microscopy, we introduce a new large scale microscopic image malaria dataset. Thirty-eight thousand cells are tagged from the 345 microscopic images of different Giemsa-stained slides of blood samples. Extensive experimentation is performed using different CNN backbones including VGG, DenseNet, and ResNet on this dataset. Our experiments and analysis reveal that the two-stage approach works better than the one-stage approach for malaria detection. To ensure the usability of our approach, we have also developed a mobile app that will be used by local hospitals for investigation and educational purposes. The dataset, its annotations, and implementation codes will be released upon publication of the paper.

Self-supervised learning approaches for unsupervised domain adaptation (UDA) of semantic segmentation models suffer from challenges of predicting and selecting reasonable good quality pseudo labels. In this paper, we propose a novel approach of exploiting scale-invariance property of the semantic segmentation model for self-supervised domain adaptation. Our algorithm is based on a reasonable assumption that, in general, regardless of the size of the object and stuff (given context) the semantic labeling should be unchanged. We show that this constraint is violated over the images of the target domain, and hence could be used to transfer labels in-between differently scaled patches. Specifically, we show that semantic segmentation model produces output with high entropy when presented with scaled-up patches of target domain, in comparison to when presented original size images. These scale-invariant examples are extracted from the most confident images of the target domain. Dynamic class specific entropy thresholding mechanism is presented to filter out unreliable pseudo-labels. Furthermore, we also incorporate the focal loss to tackle the problem of class imbalance in self-supervised learning. Extensive experiments have been performed, and results indicate that exploiting the scale-invariant labeling, we outperform existing self-supervised based state-of-the-art domain adaptation methods. Specifically, we achieve 1.3% and 3.8% of lead for GTA5 to Cityscapes and SYNTHIA to Cityscapes with VGG16-FCN8 baseline network., Comment: Accepted to ECCV2020

This paper proposes a novel domain adaptation algorithm to handle the challenges posed by the satellite and aerial imagery, and demonstrates its effectiveness on the built-up region segmentation problem. Built-up area estimation is an important component in understanding the human impact on the environment, the effect of public policy, and general urban population analysis. The diverse nature of aerial and satellite imagery and lack of labeled data covering this diversity makes machine learning algorithms difficult to generalize for such tasks, especially across multiple domains. On the other hand, due to the lack of strong spatial context and structure, in comparison to the ground imagery, the application of existing unsupervised domain adaptation methods results in the sub-optimal adaptation. We thoroughly study the limitations of existing domain adaptation methods and propose a weakly-supervised adaptation strategy where we assume image-level labels are available for the target domain. More specifically, we design a built-up area segmentation network (as encoder-decoder), with an image classification head added to guide the adaptation. The devised system is able to address the problem of visual differences in multiple satellite and aerial imagery datasets, ranging from high resolution (HR) to very high resolution (VHR). A realistic and challenging HR dataset is created by hand-tagging the 73.4 sq-km of Rwanda, capturing a variety of build-up structures over different terrain. The developed dataset is spatially rich compared to existing datasets and covers diverse built-up scenarios including built-up areas in forests and deserts, mud houses, tin, and colored rooftops. Extensive experiments are performed by adapting from the single-source domain, to segment out the target domain. We achieve high gains ranging 11.6%-52% in IoU over the existing state-of-the-art methods., Comment: Accepted at ISPRS Journal of Photogrammetry and Remote Sensing

Visual identification of gunmen in a crowd is a challenging problem, that requires resolving the association of a person with an object (firearm). We present a novel approach to address this problem, by defining human-object interaction (and non-interaction) bounding boxes. In a given image, human and firearms are separately detected. Each detected human is paired with each detected firearm, allowing us to create a paired bounding box that contains both object and the human. A network is trained to classify these paired-bounding-boxes into human carrying the identified firearm or not. Extensive experiments were performed to evaluate effectiveness of the algorithm, including exploiting full pose of the human, hand key-points, and their association with the firearm. The knowledge of spatially localized features is key to success of our method by using multi-size proposals with adaptive average pooling. We have also extended a previously firearm detection dataset, by adding more images and tagging in extended dataset the human-firearm pairs (including bounding boxes for firearms and gunmen). The experimental results ($AP_{hold} = 78.5$) demonstrate effectiveness of the proposed method., Comment: 5 pages, accepted in IEEE ICIP 2020