Prevalence of voxel-based 3D single-stage detectors contrast with underexplored point-based methods. In this paper, we present a lightweight single stage object detector 3DSSD to achieve decent balance accuracy and efficiency. paradigm, all upsampling layers the refinement stage, which are indispensable in existing methods, abandoned. We instead propose fusion sampling strategy downsampling process make detection on less representative points feasible. A delicate box prediction network,...

We propose a two-stage 3D object detection framework, named sparse-to-dense Object Detector (STD). The first stage is bottom-up proposal generation network that uses raw point clouds as input to generate accurate proposals by seeding each with new spherical anchor. It achieves higher recall less computation compared prior works. Then, PointsPool applied for feature transforming interior features from sparse expression compact representation, which saves even more computation. In box...

Convolutional Neural Networks (CNNs) have gained a remarkable success on many image classification tasks in recent years. However, the performance of CNNs highly relies upon their architectures. For most state-of-the-art CNNs, architectures are often manually-designed with expertise both and investigated problems. Therefore, it is difficult for users, who no extended to design optimal CNN own problems interest. In this paper, we propose an automatic architecture method by using genetic...

Evolutionary paradigms have been successfully applied to neural network designs for two decades. Unfortunately, these methods cannot scale well the modern deep networks due complicated architectures and large quantities of connection weights. In this paper, we propose a new method using genetic algorithms evolving weight initialization values convolutional address image classification problems. proposed algorithm, an efficient variable-length gene encoding strategy is designed represent...

Inverted generational distance (IGD) has been widely considered as a reliable performance indicator to concurrently quantify the convergence and diversity of multiobjective many-objective evolutionary algorithms. In this paper, an IGD indicator-based algorithm for solving optimization problems (MaOPs) proposed. Specifically, is employed in each generation select solutions with favorable diversity. addition, computationally efficient dominance comparison method designed assign rank values...

Deep Neural Networks (DNNs) have achieved great success in many applications. The architectures of DNNs play a crucial role their performance, which is usually manually designed with rich expertise. However, such design process labour intensive because the trial-and-error process, and also not easy to realize due rare expertise practice. Architecture Search (NAS) type technology that can automatically. Among different methods NAS, Evolutionary Computation (EC) recently gained much attention...

The performance of convolutional neural networks (CNNs) highly relies on their architectures. In order to design a CNN with promising performance, extensive expertise in both CNNs and the investigated problem domain is required, which not necessarily available every interested user. To address this problem, we propose automatically evolve architectures by using genetic algorithm (GA) based ResNet DenseNet blocks. proposed completely automatic designing particular, neither preprocessing...

Convolutional neural networks (CNNs) have shown remarkable performance in various real-world applications. Unfortunately, the promising of CNNs can be achieved only when their architectures are optimally constructed. The state-of-the-art typically handcrafted with extensive expertise both and investigated data, which consequently hampers widespread adoption for less experienced users. Evolutionary deep learning (EDL) is able to automatically design best CNN without much expertise. However,...

Artificial intelligence (AI) emphasises the creation of intelligent machines/systems that function like humans. AI has been applied to many real-world applications. Machine learning is a branch based on idea systems can learn from data, identify hidden patterns, and make decisions with little/minimal human intervention. Evolutionary computation an umbrella population-based intelligent/learning algorithms inspired by nature, where New Zealand good international reputation. This paper provides...

Deep Learning (DL) aims at learning the \emph{meaningful representations}. A meaningful representation refers to one that gives rise significant performance improvement of associated Machine (ML) tasks by replacing raw data as input. However, optimal architecture design and model parameter estimation in DL algorithms are widely considered be intractable. Evolutionary much preferable for complex non-convex problems due its inherent characteristics gradient-free insensitivity local optimum. In...

Convolutional neural networks (CNNs) are one of the most effective deep learning methods to solve image classification problems, but best architecture a CNN specific problem can be extremely complicated and hard design. This paper focuses on utilising Particle Swarm Optimisation (PSO) automatically search for optimal CNNs without any manual work involved. In order achieve goal, three improvements made based traditional PSO. First, novel encoding strategy inspired by computer which empowers...

Convolutional auto-encoders have shown their remarkable performance in stacking to deep convolutional neural networks for classifying image data during past several years. However, they are unable construct the state-of-the-art due intrinsic architectures. In this regard, we propose a flexible auto-encoder by eliminating constraints on numbers of layers and pooling from traditional auto-encoder. We also design an architecture discovery method using particle swarm optimization, which is...

We present a novel 3D object detection framework, named IPOD, based on raw point cloud. It seeds proposal for each point, which is the basic element. This paradigm provides us with high recall and fidelity of information, leading to suitable way process cloud data. design an end-to-end trainable architecture, where features all points within are extracted from backbone network achieve feature final bounding inference. These both context information precise coordinates yield improved...

Convergence and diversity are interdependently handled during the evolutionary process by most existing many-objective algorithms (MaOEAs). In such a design, degraded performance of one would deteriorate other, only solutions with both able to improve MaOEAs. Unfortunately, it is not easy constantly maintain population convergence diversity. this paper, an MaOEA based on two independent stages proposed for effectively solving optimization problems (MaOPs), where addressed in sequential...

Natural image matting separates the foreground from background in fractional occupancy which can be caused by highly transparent objects, complex (e.g., net or tree), and/or objects containing very fine details hairs). Although conventional formulation applied to all of above cases, no previous work has attempted reason underlying causes due various semantics.We show how obtain better alpha mattes incorporating into our framework semantic classification regions. Specifically, we consider and...

Currently, there have been many kinds of voxel-based 3D single stage detectors, while point-based methods are still underexplored. In this paper, we first present a lightweight and effective object detector, named 3DSSD, achieving good balance between accuracy efficiency. paradigm, all upsampling layers refinement stage, which indispensable in existing methods, abandoned to reduce the large computation cost. We novelly propose fusion sampling strategy downsampling process make detection on...

In recent years, convolutional neural networks (CNNs) have become deeper in order to achieve better classification accuracy image classification. However, it is difficult deploy the state-of-the-art deep CNNs for industrial use due difficulty of manually fine-tuning hyperparameters and trade-off between computational cost. This paper proposes a novel multi-objective optimization method evolving real-life applications, which automatically evolves non-dominant solutions at Pareto front. Three...

Despite the significant progress made by deep learning in natural image matting, there has been so far no representative work on for video matting due to inherent technical challenges reasoning temporal domain and lack of large-scale datasets. In this paper, we propose a learning-based framework which employs novel effective spatio-temporal feature aggregation module (ST-FAM). As optical flow estimation can be very unreliable within regions, ST-FAM is designed effectively align aggregate...

Evolutionary neural architecture search (ENAS) can automatically design the architectures of deep networks (DNNs) using evolutionary computation algorithms. However, most ENAS algorithms require an intensive computational resource, which is not necessarily available to users interested. Performance predictors are a type regression models assist accomplish search, while without exerting much resource. Despite various performance have been designed, they employ same training protocol build...

Background: Exendin-4 is an incretin mimetic agent approved for type 2 diabetes treatment. However, the required frequent injections restrict its clinical application. Here, potential use of chitosan-coated poly (d,l-lactide-co-glycolide) (CS-PLGA) nanoparticles was investigated intestinal delivery exendin-4. Methods and results: Nanoparticles were prepared using a modified water–oil–water (w/o/w) emulsion solvent-evaporation method, followed by coating with chitosan. The physical...

We present a new two-stage 3D object detection framework, named sparse-to-dense Object Detector (STD). The first stage is bottom-up proposal generation network that uses raw point cloud as input to generate accurate proposals by seeding each with spherical anchor. It achieves high recall less computation compared prior works. Then, PointsPool applied for generating features transforming their interior from sparse expression compact representation, which saves even more time. In box...