Decode-MOT: How Can We Hurdle Frames to Go Beyond Tracking-by-Detection?

The speed of tracking-by-detection (TBD) greatly depends on the number of running a detector because the detection is the most expensive operation in TBD. In many practical cases, multi-object tracking (MOT) can be, however, achieved based tracking-by-motion (TBM) only. This is a possible solution without much loss of MOT accuracy when the variations of object cardinality and motions are not much within consecutive frames. Therefore, the MOT problem can be transformed to find the best TBD and TBM mechanism. To achieve it, we propose a novel decision coordinator for MOT (Decode-MOT) which can determine the best TBD/TBM mechanism according to scene and tracking contexts. In specific, our Decode-MOT learns tracking and scene contextual similarities between frames. Because the contextual similarities can vary significantly according to the used trackers and tracking scenes, we learn the Decode-MOT via self-supervision. The evaluation results on MOT challenge datasets prove that our method can boost the tracking speed greatly while keeping the state-of-the-art MOT accuracy. Our code will be available at https://github.com/reussite-cv/Decode-MOT.

Deformable Part Region Learning and Feature Aggregation Tree Representation for Object Detection.

Region-based object detection infers object regions for one or more categories in an image. Due to the recent advances in deep learning and region proposal methods, object detectors based on convolutional neural networks (CNNs) have been flourishing and provided promising detection results. However, the accuracy of the convolutional object detectors can be degraded often due to the low feature discriminability caused by geometric variation or transformation of an object. In this article, we propose a deformable part region (DPR) learning in order to allow decomposed part regions to be deformable according to the geometric transformation of an object. Because the ground truth of the part models is not available in many cases, we design part model losses for the detection and segmentation, and learn the geometric parameters by minimizing an integral loss including those part losses. As a result, we can train our DPR network without extra supervision, and make multi-part models deformable according to object geometric variation. Moreover, we propose a novel feature aggregation tree (FAT) so as to learn more discriminative region of interest (RoI) features via bottom-up tree construction. The FAT can learn the stronger semantic features by aggregating part RoI features along the bottom-up pathways of the tree. We also present a spatial and channel attention mechanism for the aggregation between different node features. Based on the proposed DPR and FAT networks, we design a new cascade architecture that can refine detection tasks iteratively. Without bells and whistles, we achieve impressive detection and segmentation results on MSCOCO and PASCAL VOC datasets. Our Cascade D-PRD achieves the 57.9 box AP with the Swin-L backbone. We also provide an extensive ablation study to prove the effectiveness and usefulness of the proposed methods for large-scale object detection.

Effective Multi-Object Tracking via Global Object Models and Object Constraint Learning.

Effective multi-object tracking is still challenging due to the trade-off between tracking accuracy and speed. Because the recent multi-object tracking (MOT) methods leverage object appearance and motion models so as to associate detections between consecutive frames, the key for effective multi-object tracking is to reduce the computational complexity of learning both models. To this end, this work proposes global appearance and motion models to discriminate multiple objects instead of learning local object-specific models. In concrete detail, it learns a global appearance model using contrastive learning between object appearances. In addition, we learn a global relation motion model using relative motion learning between objects. Moreover, this paper proposes object constraint learning for improving tracking efficiency. This study considers the discriminability of the models as a constraint, and learns both models when inconsistency with the constraint occurs. Therefore, object constraint learning differs from the conventional online learning for multi-object tracking which updates learnable parameters per frame. This work incorporates global models and object constraint learning into the confidence-based association method, and compare our tracker with the state-of-the-art methods on public available MOT Challenge datasets. As a result, we achieve 64.5% MOTA (multi-object tracking accuracy) and 6.54 Hz tracking speed on the MOT16 test dataset. The comparison results show that our methods can contribute to improve tracking accuracy and tracking speed together.

Configurationally Random Polythiophene for Improved Polymer Ordering and Charge-Transporting Ability.

Random polythiophene polymers are characterized by the arbitrary sequences of monomeric units along polymer backbones. These untailored orientations generally result in the twisting of thiophene rings out of the conjugation planarity in addition to steric repulsions experienced among substituted alkyl chains. These tendencies have limited close polymer packing, which has been detrimental to charge transport in these moieties. To ameliorate charge transport in these classes of polymers, we make use of simple Stille coupling polymerization to synthesize highly random polythiophene polymers. We induced a positive microstructural change between polymer chains by attuning the ratio between alkyl-substituted and nonalkyl-substituted monomer units along the backbones. The optimized random polythiophene was found to have enhanced intermolecular interaction, increased size of crystallites, and stronger tendency to take edge orientation compared with both regiorandom and regioregular poly(3-hexylthiophene) polymers. Incorporation of the optimized random polythiophene as an active material in solid-state electrolyte-gated organic field-effect transistors exhibited better performance than the control device using regioregular poly(3-hexylthiophene), with a high hole mobility up to 4.52 cm2 V-1 s-1 in ambient conditions.

Electrical Transport and Thermoelectric Properties of SnSe-SnTe Solid Solution.

SnSe is considered as a promising thermoelectric (TE) material since the discovery of the record figure of merit (ZT) of 2.6 at 926 K in single crystal SnSe. It is, however, difficult to use single crystal SnSe for practical applications due to the poor mechanical properties and the difficulty and cost of fabricating a single crystal. It is highly desirable to improve the properties of polycrystalline SnSe whose TE properties are still not near to that of single crystal SnSe. In this study, in order to control the TE properties of polycrystalline SnSe, polycrystalline SnSe-SnTe solid solutions were fabricated, and the effect of the solid solution on the electrical transport and TE properties was investigated. The SnSe1-xTex samples were fabricated using mechanical alloying and spark plasma sintering. X-ray diffraction (XRD) analyses revealed that the solubility limit of Te in SnSe1-xTex is somewhere between x = 0.3 and 0.5. With increasing Te content, the electrical conductivity was increased due to the increase of carrier concentration, while the lattice thermal conductivity was suppressed by the increased amount of phonon scattering. The change of carrier concentration and electrical conductivity is explained using the measured band gap energy and the calculated band structure. The change of thermal conductivity is explained using the change of lattice thermal conductivity from the increased amount of phonon scattering at the point defect sites. A ZT of ~0.78 was obtained at 823 K from SnSe0.7Te0.3, which is an ~11% improvement compared to that of SnSe.

3D Printer-Based Encapsulated Origami Electronics for Extreme System Stretchability and High Areal Coverage.

Stretchability and areal coverage of active devices are critical design considerations of stretchable or wearable photovoltaics and photodetections where high areal coverages are required. However, simultaneously maximizing both properties in conventional island-bridge structures through traditional two-dimensional manufacturing processes is difficult due to their inherent trade-offs. Here, a 3D printer-based strategy to achieve extreme system stretchability and high areal coverage through combining fused deposition modeling (FDM) and flexible conductive nanocomposites is reported. Distinguished from typical approaches of using conductive filaments for FDM which have a flexibility dilemma and conductivity trade-offs, the proposed axiomatic approach to embed a two-dimensional silver nanowire percolation network into the surfaces of flexible 3D printed structures offers sufficient conductivity and deformability as well as additional benefits of electrical junction enhancement and encapsulation of silver nanowires. Kirigami/origami-pattern-guided three-dimensional arrangements of encapsulated interconnections provide efficient control over stretchability and areal coverage. The suggested process enables a perovskite solar module with an initial areal coverage of ∼97% to be electrically and mechanically reversible with 400% system stretchability and 25 000% interconnect stretchability under the 1000 cycle test, by folding down or hiding the origami-applied interconnects under the islands. This 3D printing strategy of potentially low cost, large size capabilities, and high speed is promising for highly flexible future energy conversion applications.

Controlling the Morphology of Organic-Inorganic Hybrid Perovskites through Dual Additive-Mediated Crystallization for Solar Cell Applications.

To realize a high-efficiency perovskite solar cell (PSC), it is critical to optimize the morphology of the perovskite film for a uniform and smooth finish with large grain size during film formation. Using a chemical compound as an additive to the precursor solution has recently been established as a promising method to control the morphology of the perovskite film. In this study, we propose a new method to achieve an improved morphology of the methylammonium lead iodide perovskite film by simultaneous addition of dimethyl sulfoxide (DMSO) and methoxyammonium salt (MeO) (dual additives). We demonstrated that an appropriate amount of the MeO additive helps the precursors form a stable intermediated PbI2-DMSO adduct during film formation and enlarges the perovskite grains by retarding the kinetics of conversion of the adduct to the perovskite. Furthermore, we experimentally observed that the optical band gaps and crystal structures of perovskite films are reasonably unaffected by the MeO additive because MeO is almost eliminated during annealing. By optimizing the amount of MeO, we achieved improved device performances of the PSCs with a high power conversion efficiency of 19.71% that is ∼15% higher than that obtained for the control device (17.15%).

Confidence-Based Data Association and Discriminative Deep Appearance Learning for Robust Online Multi-Object Tracking.

Online multi-object tracking aims at estimating the tracks of multiple objects instantly with each incoming frame and the information provided up to the moment. It still remains a difficult problem in complex scenes, because of the large ambiguity in associating multiple objects in consecutive frames and the low discriminability between objects appearances. In this paper, we propose a robust online multi-object tracking method that can handle these difficulties effectively. We first define the tracklet confidence using the detectability and continuity of a tracklet, and decompose a multi-object tracking problem into small subproblems based on the tracklet confidence. We then solve the online multi-object tracking problem by associating tracklets and detections in different ways according to their confidence values. Based on this strategy, tracklets sequentially grow with online-provided detections, and fragmented tracklets are linked up with others without any iterative and expensive association steps. For more reliable association between tracklets and detections, we also propose a deep appearance learning method to learn a discriminative appearance model from large training datasets, since the conventional appearance learning methods do not provide rich representation that can distinguish multiple objects with large appearance variations. In addition, we combine online transfer learning for improving appearance discriminability by adapting the pre-trained deep model during online tracking. Experiments with challenging public datasets show distinct performance improvement over other state-of-the-arts batch and online tracking methods, and prove the effect and usefulness of the proposed methods for online multi-object tracking.

Solution-Processed Ultrathin TiO2 Compact Layer Hybridized with Mesoporous TiO2 for High-Performance Perovskite Solar Cells.

The electron transport layer (ETL) is a key component of perovskite solar cells (PSCs) and must provide efficient electron extraction and collection while minimizing the charge recombination at interfaces in order to ensure high performance. Conventional bilayered TiO2 ETLs fabricated by depositing compact TiO2 (c-TiO2) and mesoporous TiO2 (mp-TiO2) in sequence exhibit resistive losses due to the contact resistance at the c-TiO2/mp-TiO2 interface and the series resistance arising from the intrinsically low conductivity of TiO2. Herein, to minimize such resistive losses, we developed a novel ETL consisting of an ultrathin c-TiO2 layer hybridized with mp-TiO2, which is fabricated by performing one-step spin-coating of a mp-TiO2 solution containing a small amount of titanium diisopropoxide bis(acetylacetonate) (TAA). By using electron microscopies and elemental mapping analysis, we establish that the optimal concentration of TAA produces an ultrathin blocking layer with a thickness of ∼3 nm and ensures that the mp-TiO2 layer has a suitable porosity for efficient perovskite infiltration. We compare PSCs based on mesoscopic ETLs with and without compact layers to determine the role of the hole-blocking layer in their performances. The hybrid ETLs exhibit enhanced electron extraction and reduced charge recombination, resulting in better photovoltaic performances and reduced hysteresis of PSCs compared to those with conventional bilayered ETLs.

Performance enhancement of planar heterojunction perovskite solar cells by n-doping of the electron transporting layer.

Herein we report a simple n-doping method to enhance the performance of perovskite solar cells with a planar heterojunction structure. Devices with an n-doped PCBM electron transporting layer exhibit a power conversion efficiency of 13.8% with a remarkably enhanced short-circuit current of 22.0 mA cm(-2) as compared to the devices with an un-doped PCBM layer.

Polyp Detection via Imbalanced Learning and Discriminative Feature Learning.

Recent achievement of the learning-based classification leads to the noticeable performance improvement in automatic polyp detection. Here, building large good datasets is very crucial for learning a reliable detector. However, it is practically challenging due to the diversity of polyp types, expensive inspection, and labor-intensive labeling tasks. For this reason, the polyp datasets usually tend to be imbalanced, i.e., the number of non-polyp samples is much larger than that of polyp samples, and learning with those imbalanced datasets results in a detector biased toward a non-polyp class. In this paper, we propose a data sampling-based boosting framework to learn an unbiased polyp detector from the imbalanced datasets. In our learning scheme, we learn multiple weak classifiers with the datasets rebalanced by up/down sampling, and generate a polyp detector by combining them. In addition, for enhancing discriminability between polyps and non-polyps that have similar appearances, we propose an effective feature learning method using partial least square analysis, and use it for learning compact and discriminative features. Experimental results using challenging datasets show obvious performance improvement over other detectors. We further prove effectiveness and usefulness of the proposed methods with extensive evaluation.

Robust online multiobject tracking with data association and track management.

In this paper, we consider a multiobject tracking problem in complex scenes. Unlike batch tracking systems using detections of the entire sequence, we propose a novel online multiobject tracking system in order to build tracks sequentially using online provided detections. To track objects robustly even under frequent occlusions, the proposed system consists of three main parts: 1) visual tracking with a novel data association with a track existence probability by associating online detections with the corresponding tracks under partial occlusions; 2) track management to associate terminated tracks for linking tracks fragmented by long-term occlusions; and 3) online model learning to generate discriminative appearance models for successful associations in other two parts. Experimental results using challenging public data sets show the obvious performance improvement of the proposed system, compared with other state-of-the-art tracking systems. Furthermore, extensive performance analysis of the three main parts demonstrates effects and usefulness of the each component for multiobject tracking.