Search

In recent years, transformer-based architectures become the de facto standard for sequence modeling in deep learning frameworks. Inspired by the successful examples, we propose a causal visual-inertial fusion transformer (VIFT) for pose estimation in deep visual-inertial odometry. This study aims to improve pose estimation accuracy by leveraging the attention mechanisms in transformers, which better utilize historical data compared to the recurrent neural network (RNN) based methods seen in recent methods. Transformers typically require large-scale data for training. To address this issue, we utilize inductive biases for deep VIO networks. Since latent visual-inertial feature vectors encompass essential information for pose estimation, we employ transformers to refine pose estimates by updating latent vectors temporally. Our study also examines the impact of data imbalance and rotation learning methods in supervised end-to-end learning of visual inertial odometry by utilizing specialized gradients in backpropagation for the elements of SE$(3)$ group. The proposed method is end-to-end trainable and requires only a monocular camera and IMU during inference. Experimental results demonstrate that VIFT increases the accuracy of monocular VIO networks, achieving state-of-the-art results when compared to previous methods on the KITTI dataset. The code will be made available at https://github.com/ybkurt/VIFT., Comment: Accepted to ECCV 2024 2nd Workshop on Vision-Centric Autonomous Driving (VCAD)

3D Gaussian Splatting has recently shown promising results as an alternative scene representation in SLAM systems to neural implicit representations. However, current methods either lack dense depth maps to supervise the mapping process or detailed training designs that consider the scale of the environment. To address these drawbacks, we present IG-SLAM, a dense RGB-only SLAM system that employs robust Dense-SLAM methods for tracking and combines them with Gaussian Splatting. A 3D map of the environment is constructed using accurate pose and dense depth provided by tracking. Additionally, we utilize depth uncertainty in map optimization to improve 3D reconstruction. Our decay strategy in map optimization enhances convergence and allows the system to run at 10 fps in a single process. We demonstrate competitive performance with state-of-the-art RGB-only SLAM systems while achieving faster operation speeds. We present our experiments on the Replica, TUM-RGBD, ScanNet, and EuRoC datasets. The system achieves photo-realistic 3D reconstruction in large-scale sequences, particularly in the EuRoC dataset., Comment: 8 pages, 3 page ref, 5 figures

We introduce, XoFTR, a cross-modal cross-view method for local feature matching between thermal infrared (TIR) and visible images. Unlike visible images, TIR images are less susceptible to adverse lighting and weather conditions but present difficulties in matching due to significant texture and intensity differences. Current hand-crafted and learning-based methods for visible-TIR matching fall short in handling viewpoint, scale, and texture diversities. To address this, XoFTR incorporates masked image modeling pre-training and fine-tuning with pseudo-thermal image augmentation to handle the modality differences. Additionally, we introduce a refined matching pipeline that adjusts for scale discrepancies and enhances match reliability through sub-pixel level refinement. To validate our approach, we collect a comprehensive visible-thermal dataset, and show that our method outperforms existing methods on many benchmarks., Comment: CVPR Image Matching Workshop, 2024. 12 pages, 7 figures, 5 tables. Codes and dataset are available at https://github.com/OnderT/XoFTR

Typical technique in knowledge distillation (KD) is regularizing the learning of a limited capacity model (student) by pushing its responses to match a powerful model's (teacher). Albeit useful especially in the penultimate layer and beyond, its action on student's feature transform is rather implicit, limiting its practice in the intermediate layers. To explicitly embed the teacher's knowledge in feature transform, we propose a learnable KD layer for the student which improves KD with two distinct abilities: i) learning how to leverage the teacher's knowledge, enabling to discard nuisance information, and ii) feeding forward the transferred knowledge deeper. Thus, the student enjoys the teacher's knowledge during the inference besides training. Formally, we repurpose 1x1-BN-ReLU-1x1 convolution block to assign a semantic vector to each local region according to the template (supervised by the teacher) that the corresponding region of the student matches. To facilitate template learning in the intermediate layers, we propose a novel form of supervision based on the teacher's decisions. Through rigorous experimentation, we demonstrate the effectiveness of our approach on 3 popular classification benchmarks. Code is available at: https://github.com/adagorgun/letKD-framework, Comment: Accepted at the British Machine Vision Conference 2023 (BMVC 2023)

A common architectural choice for deep metric learning is a convolutional neural network followed by global average pooling (GAP). Albeit simple, GAP is a highly effective way to aggregate information. One possible explanation for the effectiveness of GAP is considering each feature vector as representing a different semantic entity and GAP as a convex combination of them. Following this perspective, we generalize GAP and propose a learnable generalized sum pooling method (GSP). GSP improves GAP with two distinct abilities: i) the ability to choose a subset of semantic entities, effectively learning to ignore nuisance information, and ii) learning the weights corresponding to the importance of each entity. Formally, we propose an entropy-smoothed optimal transport problem and show that it is a strict generalization of GAP, i.e., a specific realization of the problem gives back GAP. We show that this optimization problem enjoys analytical gradients enabling us to use it as a direct learnable replacement for GAP. We further propose a zero-shot loss to ease the learning of GSP. We show the effectiveness of our method with extensive evaluations on 4 popular metric learning benchmarks. Code is available at: GSP-DML Framework, Comment: Accepted as a conference paper at International Conference on Computer Vision (ICCV) 2023

Global average pooling (GAP) is a popular component in deep metric learning (DML) for aggregating features. Its effectiveness is often attributed to treating each feature vector as a distinct semantic entity and GAP as a combination of them. Albeit substantiated, such an explanation's algorithmic implications to learn generalizable entities to represent unseen classes, a crucial DML goal, remain unclear. To address this, we formulate GAP as a convex combination of learnable prototypes. We then show that the prototype learning can be expressed as a recursive process fitting a linear predictor to a batch of samples. Building on that perspective, we consider two batches of disjoint classes at each iteration and regularize the learning by expressing the samples of a batch with the prototypes that are fitted to the other batch. We validate our approach on 4 popular DML benchmarks., Comment: \c{opyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

Utilization of event-based cameras is expected to improve the visual quality of video frame interpolation solutions. We introduce a learning-based method to exploit moving region boundaries in a video sequence to increase the overall interpolation quality.Event cameras allow us to determine moving areas precisely; and hence, better video frame interpolation quality can be achieved by emphasizing these regions using an appropriate loss function. The results show a notable average \textit{PSNR} improvement of $1.3$ dB for the tested data sets, as well as subjectively more pleasing visual results with less ghosting and blurry artifacts., Comment: Submitted to International Conference on Image Processing (ICIP) 2023

Convolution blocks serve as local feature extractors and are the key to success of the neural networks. To make local semantic feature embedding rather explicit, we reformulate convolution blocks as feature selection according to the best matching kernel. In this manner, we show that typical ResNet blocks indeed perform local feature embedding via template matching once batch normalization (BN) followed by a rectified linear unit (ReLU) is interpreted as arg-max optimizer. Following this perspective, we tailor a residual block that explicitly forces semantically meaningful local feature embedding through using label information. Specifically, we assign a feature vector to each local region according to the classes that the corresponding region matches. We evaluate our method on three popular benchmark datasets with several architectures for image classification and consistently show that our approach substantially improves the performance of the baseline architectures., Comment: Accepted at the British Machine Vision Conference 2022 (BMVC 2022)

Video frame interpolation (VFI) is a fundamental vision task that aims to synthesize several frames between two consecutive original video images. Most algorithms aim to accomplish VFI by using only keyframes, which is an ill-posed problem since the keyframes usually do not yield any accurate precision about the trajectories of the objects in the scene. On the other hand, event-based cameras provide more precise information between the keyframes of a video. Some recent state-of-the-art event-based methods approach this problem by utilizing event data for better optical flow estimation to interpolate for video frame by warping. Nonetheless, those methods heavily suffer from the ghosting effect. On the other hand, some of kernel-based VFI methods that only use frames as input, have shown that deformable convolutions, when backed up with transformers, can be a reliable way of dealing with long-range dependencies. We propose event-based video frame interpolation with attention (E-VFIA), as a lightweight kernel-based method. E-VFIA fuses event information with standard video frames by deformable convolutions to generate high quality interpolated frames. The proposed method represents events with high temporal resolution and uses a multi-head self-attention mechanism to better encode event-based information, while being less vulnerable to blurring and ghosting artifacts; thus, generating crispier frames. The simulation results show that the proposed technique outperforms current state-of-the-art methods (both frame and event-based) with a significantly smaller model size., Comment: Accepted to 2023 IEEE International Conference on Robotics and Automation (ICRA 2023)

Deep metric learning (DML) aims to minimize empirical expected loss of the pairwise intra-/inter- class proximity violations in the embedding space. We relate DML to feasibility problem of finite chance constraints. We show that minimizer of proxy-based DML satisfies certain chance constraints, and that the worst case generalization performance of the proxy-based methods can be characterized by the radius of the smallest ball around a class proxy to cover the entire domain of the corresponding class samples, suggesting multiple proxies per class helps performance. To provide a scalable algorithm as well as exploiting more proxies, we consider the chance constraints implied by the minimizers of proxy-based DML instances and reformulate DML as finding a feasible point in intersection of such constraints, resulting in a problem to be approximately solved by iterative projections. Simply put, we repeatedly train a regularized proxy-based loss and re-initialize the proxies with the embeddings of the deliberately selected new samples. We applied our method with 4 well-accepted DML losses and show the effectiveness with extensive evaluations on 4 popular DML benchmarks. Code is available at: https://github.com/yetigurbuz/ccp-dml, Comment: Accepted as a conference paper at IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2024

Hard example mining methods generally improve the performance of the object detectors, which suffer from imbalanced training sets. In this work, two existing hard example mining approaches (LRM and focal loss, FL) are adapted and combined in a state-of-the-art real-time object detector, YOLOv5. The effectiveness of the proposed approach for improving the performance on hard examples is extensively evaluated. The proposed method increases mAP by 3% compared to using the original loss function and around 1-2% compared to using the hard-mining methods (LRM or FL) individually on 2021 Anti-UAV Challenge Dataset., Comment: ICIP 2022. 5 pages, 2 figures, 7 tables. The codes are available at https://github.com/aybora/yolov5Loss

Deep learning-based image matching methods are improved significantly during the recent years. Although these methods are reported to outperform the classical techniques, the performance of the classical methods is not examined in detail. In this study, we compare classical and learning-based methods by employing mutual nearest neighbor search with ratio test and optimizing the ratio test threshold to achieve the best performance on two different performance metrics. After a fair comparison, the experimental results on HPatches dataset reveal that the performance gap between classical and learning-based methods is not that significant. Throughout the experiments, we demonstrated that SuperGlue is the state-of-the-art technique for the image matching problem on HPatches dataset. However, if a single parameter, namely ratio test threshold, is carefully optimized, a well-known traditional method SIFT performs quite close to SuperGlue and even outperforms in terms of mean matching accuracy (MMA) under 1 and 2 pixel thresholds. Moreover, a recent approach, DFM, which only uses pre-trained VGG features as descriptors and ratio test, is shown to outperform most of the well-trained learning-based methods. Therefore, we conclude that the parameters of any classical method should be analyzed carefully before comparing against a learning-based technique., Comment: 8 pages, 2 figures, 3 tables, ICCV 2021 TradiCV Workshop

A novel image matching method is proposed that utilizes learned features extracted by an off-the-shelf deep neural network to obtain a promising performance. The proposed method uses pre-trained VGG architecture as a feature extractor and does not require any additional training specific to improve matching. Inspired by well-established concepts in the psychology area, such as the Mental Rotation paradigm, an initial warping is performed as a result of a preliminary geometric transformation estimate. These estimates are simply based on dense matching of nearest neighbors at the terminal layer of VGG network outputs of the images to be matched. After this initial alignment, the same approach is repeated again between reference and aligned images in a hierarchical manner to reach a good localization and matching performance. Our algorithm achieves 0.57 and 0.80 overall scores in terms of Mean Matching Accuracy (MMA) for 1 pixel and 2 pixels thresholds respectively on Hpatches dataset, which indicates a better performance than the state-of-the-art., Comment: CVPR 2021 Image Matching Workshop Camera Ready Version

We propose a semi-automatic bounding box annotation method for visual object tracking by utilizing temporal information with a tracking-by-detection approach. For detection, we use an off-the-shelf object detector which is trained iteratively with the annotations generated by the proposed method, and we perform object detection on each frame independently. We employ Multiple Hypothesis Tracking (MHT) to exploit temporal information and to reduce the number of false-positives which makes it possible to use lower objectness thresholds for detection to increase recall. The tracklets formed by MHT are evaluated by human operators to enlarge the training set. This novel incremental learning approach helps to perform annotation iteratively. The experiments performed on AUTH Multidrone Dataset reveal that the annotation workload can be reduced up to 96% by the proposed approach., Comment: Accepted to The 2nd Anti-UAV Workshop & Challenge - ICCV Workshops, 2021. Resulting uav_detection_2 annotations and our codes are publicly available at https://github.com/aybora/Semi-Automatic-Video-Annotation-OGAM

We propose an end-to-end learning approach to address deinterleaving of patterns in time series, in particular, radar signals. We link signal clustering problem to min-cost flow as an equivalent problem once the proper costs exist. We formulate a bi-level optimization problem involving min-cost flow as a sub-problem to learn such costs from the supervised training data. We then approximate the lower level optimization problem by self-attention based neural networks and provide a trainable framework that clusters the patterns in the input as the distinct flows. We evaluate our method with extensive experiments on a large dataset with several challenging scenarios to show the efficiency., Comment: 4 pages, 2 figures, 1 table

In this work, we combine 3D convolution with late temporal modeling for action recognition. For this aim, we replace the conventional Temporal Global Average Pooling (TGAP) layer at the end of 3D convolutional architecture with the Bidirectional Encoder Representations from Transformers (BERT) layer in order to better utilize the temporal information with BERT's attention mechanism. We show that this replacement improves the performances of many popular 3D convolution architectures for action recognition, including ResNeXt, I3D, SlowFast and R(2+1)D. Moreover, we provide the-state-of-the-art results on both HMDB51 and UCF101 datasets with 85.10% and 98.69% top-1 accuracy, respectively. The code is publicly available., Comment: Presented on the 2nd Workshop on Video Turing Test: Toward Human-Level Video Story Understanding, ECCV 2020

Following the recent advances in deep networks, object detection and tracking algorithms with deep learning backbones have been improved significantly; however, this rapid development resulted in the necessity of large amounts of annotated labels. Even if the details of such semi-automatic annotation processes for most of these datasets are not known precisely, especially for the video annotations, some automated labeling processes are usually employed. Unfortunately, such approaches might result with erroneous annotations. In this work, different types of annotation errors for object detection problem are simulated and the performance of a popular state-of-the-art object detector, YOLOv3, with erroneous annotations during training and testing stages is examined. Moreover, some inevitable annotation errors in CVPR-2020 Anti-UAV Challenge dataset is also examined in this manner, while proposing a solution to correct such annotation errors of this valuable data set., Comment: Best Paper Award at The 1st Anti-UAV Workshop & Challenge - CVPR Workshops, 2020

Recognition of objects with subtle differences has been used in many practical applications, such as car model recognition and maritime vessel identification. For discrimination of the objects in fine-grained detail, we focus on deep embedding learning by using a multi-task learning framework, in which the hierarchical labels (coarse and fine labels) of the samples are utilized both for classification and a quadruplet-based loss function. In order to improve the recognition strength of the learned features, we present a novel feature selection method specifically designed for four training samples of a quadruplet. By experiments, it is observed that the selection of very hard negative samples with relatively easy positive ones from the same coarse and fine classes significantly increases some performance metrics in a fine-grained dataset when compared to selecting the quadruplet samples randomly. The feature embedding learned by the proposed method achieves favorable performance against its state-of-the-art counterparts., Comment: 6 pages, 2 figures, accepted by IEEE ICIP 2019

During the training of networks for distance metric learning, minimizers of the typical loss functions can be considered as "feasible points" satisfying a set of constraints imposed by the training data. To this end, we reformulate distance metric learning problem as finding a feasible point of a constraint set where the embedding vectors of the training data satisfy desired intra-class and inter-class proximity. The feasible set induced by the constraint set is expressed as the intersection of the relaxed feasible sets which enforce the proximity constraints only for particular samples (a sample from each class) of the training data. Then, the feasible point problem is to be approximately solved by performing alternating projections onto those feasible sets. Such an approach introduces a regularization term and results in minimizing a typical loss function with a systematic batch set construction where these batches are constrained to contain the same sample from each class for a certain number of iterations. Moreover, these particular samples can be considered as the class representatives, allowing efficient utilization of hard class mining during batch construction. The proposed technique is applied with the well-accepted losses and evaluated on Stanford Online Products, CAR196 and CUB200-2011 datasets for image retrieval and clustering. Outperforming state-of-the-art, the proposed approach consistently improves the performance of the integrated loss functions with no additional computational cost and boosts the performance further by hard negative class mining., Comment: 10 pages, 3 figures, 2 tables

During the recent years, correlation filters have shown dominant and spectacular results for visual object tracking. The types of the features that are employed in these family of trackers significantly affect the performance of visual tracking. The ultimate goal is to utilize robust features invariant to any kind of appearance change of the object, while predicting the object location as properly as in the case of no appearance change. As the deep learning based methods have emerged, the study of learning features for specific tasks has accelerated. For instance, discriminative visual tracking methods based on deep architectures have been studied with promising performance. Nevertheless, correlation filter based (CFB) trackers confine themselves to use the pre-trained networks which are trained for object classification problem. To this end, in this manuscript the problem of learning deep fully convolutional features for the CFB visual tracking is formulated. In order to learn the proposed model, a novel and efficient backpropagation algorithm is presented based on the loss function of the network. The proposed learning framework enables the network model to be flexible for a custom design. Moreover, it alleviates the dependency on the network trained for classification. Extensive performance analysis shows the efficacy of the proposed custom design in the CFB tracking framework. By fine-tuning the convolutional parts of a state-of-the-art network and integrating this model to a CFB tracker, which is the top performing one of VOT2016, 18% increase is achieved in terms of expected average overlap, and tracking failures are decreased by 25%, while maintaining the superiority over the state-of-the-art methods in OTB-2013 and OTB-2015 tracking datasets., Comment: Accepted version of IEEE Transactions on Image Processing

Segmentation of an object from a video is a challenging task in multimedia applications. Depending on the application, automatic or interactive methods are desired; however, regardless of the application type, efficient computation of video object segmentation is crucial for time-critical applications; specifically, mobile and interactive applications require near real-time efficiencies. In this paper, we address the problem of video segmentation from the perspective of efficiency. We initially redefine the problem of video object segmentation as the propagation of MRF energies along the temporal domain. For this purpose, a novel and efficient method is proposed to propagate MRF energies throughout the frames via bilateral filters without using any global texture, color or shape model. Recently presented bi-exponential filter is utilized for efficiency, whereas a novel technique is also developed to dynamically solve graph-cuts for varying, non-lattice graphs in general linear filtering scenario. These improvements are experimented for both automatic and interactive video segmentation scenarios. Moreover, in addition to the efficiency, segmentation quality is also tested both quantitatively and qualitatively. Indeed, for some challenging examples, significant time efficiency is observed without loss of segmentation quality., Comment: Multimedia, IEEE Transactions on (Volume:16, Issue: 5, Aug. 2014)