arXiv subjects

Shijie Li

Publications and source records attributed to Shijie Li.

At least 73 records · Page 4Linked to original sources

arXiv2022

Crop mapping in the small sample/no sample case: an approach using a two-level cascade classifier and integrating domain knowledge

Mapping crops using remote sensing technology is important for food security and land management. Machine learning-based methods has become a popular approach for crop mapping in recent years. However, the key to machine learning, acquiring ample and accurate samples, is usually time-consuming and laborious. To solve this problem, a crop mapping method in the small sample/no sample case that integrating domain knowledge and using a cascaded classification framework that combine a weak classifier learned from samples with strong features and a strong classifier trained by samples with weak feature was proposed. First, based on the domain knowledge of various crops, a low-capacity classifier such as decision tree was applied to acquire those pixels with distinctive features and complete observation sequences as "strong feature" samples. Then, to improve the representativeness of these samples, sample augmentation strategy that artificially remove the observations of "strong feature" samples according to the average valid observation proportion in target area was applied. Finally, based on the original samples and augmented samples, a large-capacity classifier such as random forest was trained for crop mapping. The method achieved an overall accuracy of 82% in the MAP crop recognition competition held by Syngenta Group, China in 2021 (third prize, ranked fourth). This method integrates domain knowledge to overcome the difficulties of sample acquisition, providing a convenient, fast and accurate solution for crop mapping.

cs.CV↗

arXiv2022

Dual Pyramid Generative Adversarial Networks for Semantic Image Synthesis

The goal of semantic image synthesis is to generate photo-realistic images from semantic label maps. It is highly relevant for tasks like content generation and image editing. Current state-of-the-art approaches, however, still struggle to generate realistic objects in images at various scales. In particular, small objects tend to fade away and large objects are often generated as collages of patches. In order to address this issue, we propose a Dual Pyramid Generative Adversarial Network (DP-GAN) that learns the conditioning of spatially-adaptive normalization blocks at all scales jointly, such that scale information is bi-directionally used, and it unifies supervision at different scales. Our qualitative and quantitative results show that the proposed approach generates images where small and large objects look more realistic compared to images generated by state-of-the-art methods.

cs.CV↗

arXiv2022

Toward An Optimal Selection of Dialogue Strategies: A Target-Driven Approach for Intelligent Outbound Robots

With the growth of the economy and society, enterprises, especially in the FinTech industry, have increasing demands of outbound calls for customers such as debt collection, marketing, anti-fraud calls, and so on. But a large amount of repetitive and mechanical work occupies most of the time of human agents, so the cost of equipment and labor for enterprises is increasing accordingly. At the same time, with the development of artificial intelligence technology in the past few decades, it has become quite common for companies to use new technologies such as Big Data and artificial intelligence to empower outbound call businesses. The intelligent outbound robot is a typical application of the artificial intelligence technology in the field of outbound call businesses. It is mainly used to communicate with customers in order to accomplish a certain target. It has the characteristics of low cost, high reuse, and easy compliance, which has attracted more attention from the industry. At present, there are two kinds of intelligent outbound robots in the industry but both of them still leave large room for improvement. One kind of them is based on a finite state machine relying on the configuration of jump conditions and corresponding nodes based on manual experience. This kind of intelligent outbound robot is also called a flow-based robot. For example, the schematic diagram of the working model of a flow-based robot for debt collection is shown in Fig.\ref{fig:label}. In each round, the robot will reply to the user with the words corresponding to each node.

cs.CL↗

arXiv2022

Improved Counting and Localization from Density Maps for Object Detection in 2D and 3D Microscopy Imaging

Object counting and localization are key steps for quantitative analysis in large-scale microscopy applications. This procedure becomes challenging when target objects are overlapping, are densely clustered, and/or present fuzzy boundaries. Previous methods producing density maps based on deep learning have reached a high level of accuracy for object counting by assuming that object counting is equivalent to the integration of the density map. However, this model fails when objects show significant overlap regarding accurate localization. We propose an alternative method to count and localize objects from the density map to overcome this limitation. Our procedure includes the following three key aspects: 1) Proposing a new counting method based on the statistical properties of the density map, 2) optimizing the counting results for those objects which are well-detected based on the proposed counting method, and 3) improving localization of poorly detected objects using the proposed counting method as prior information. Validation includes processing of microscopy data with known ground truth and comparison with other models that use conventional processing of the density map. Our results show improved performance in counting and localization of objects in 2D and 3D microscopy data. Furthermore, the proposed method is generic, considering various applications that rely on the density map approach. Our code will be released post-review.

cs.CV↗

arXiv2021

Rethinking 3D LiDAR Point Cloud Segmentation

Many point-based semantic segmentation methods have been designed for indoor scenarios, but they struggle if they are applied to point clouds that are captured by a LiDAR sensor in an outdoor environment. In order to make these methods more efficient and robust such that they can handle LiDAR data, we introduce the general concept of reformulating 3D point-based operations such that they can operate in the projection space. While we show by means of three point-based methods that the reformulated versions are between 300 and 400 times faster and achieve a higher accuracy, we furthermore demonstrate that the concept of reformulating 3D point-based operations allows to design new architectures that unify the benefits of point-based and image-based methods. As an example, we introduce a network that integrates reformulated 3D point-based operations into a 2D encoder-decoder architecture that fuses the information from different 2D scales. We evaluate the approach on four challenging datasets for semantic LiDAR point cloud segmentation and show that leveraging reformulated 3D point-based operations with 2D image-based operations achieves very good results for all four datasets.

cs.CV↗

arXiv2021

Multi-scale Interaction for Real-time LiDAR Data Segmentation on an Embedded Platform

Real-time semantic segmentation of LiDAR data is crucial for autonomously driving vehicles, which are usually equipped with an embedded platform and have limited computational resources. Approaches that operate directly on the point cloud use complex spatial aggregation operations, which are very expensive and difficult to optimize for embedded platforms. They are therefore not suitable for real-time applications with embedded systems. As an alternative, projection-based methods are more efficient and can run on embedded platforms. However, the current state-of-the-art projection-based methods do not achieve the same accuracy as point-based methods and use millions of parameters. In this paper, we therefore propose a projection-based method, called Multi-scale Interaction Network (MINet), which is very efficient and accurate. The network uses multiple paths with different scales and balances the computational resources between the scales. Additional dense interactions between the scales avoid redundant computations and make the network highly efficient. The proposed network outperforms point-based, image-based, and projection-based methods in terms of accuracy, number of parameters, and runtime. Moreover, the network processes more than 24 scans per second on an embedded platform, which is higher than the framerates of LiDAR sensors. The network is therefore suitable for autonomous vehicles.

cs.CV↗

arXiv2021

Moving Object Segmentation in 3D LiDAR Data: A Learning-based Approach Exploiting Sequential Data

The ability to detect and segment moving objects in a scene is essential for building consistent maps, making future state predictions, avoiding collisions, and planning. In this paper, we address the problem of moving object segmentation from 3D LiDAR scans. We propose a novel approach that pushes the current state of the art in LiDAR-only moving object segmentation forward to provide relevant information for autonomous robots and other vehicles. Instead of segmenting the point cloud semantically, i.e., predicting the semantic classes such as vehicles, pedestrians, roads, etc., our approach accurately segments the scene into moving and static objects, i.e., also distinguishing between moving cars vs. parked cars. Our proposed approach exploits sequential range images from a rotating 3D LiDAR sensor as an intermediate representation combined with a convolutional neural network and runs faster than the frame rate of the sensor. We compare our approach to several other state-of-the-art methods showing superior segmentation quality in urban environments. Additionally, we created a new benchmark for LiDAR-based moving object segmentation based on SemanticKITTI. We published it to allow other researchers to compare their approaches transparently and we furthermore published our code.

cs.RO↗

arXiv2021

MiniSeg: An Extremely Minimum Network for Efficient COVID-19 Segmentation

The rapid spread of the new pandemic, i.e., COVID-19, has severely threatened global health. Deep-learning-based computer-aided screening, e.g., COVID-19 infected CT area segmentation, has attracted much attention. However, the publicly available COVID-19 training data are limited, easily causing overfitting for traditional deep learning methods that are usually data-hungry with millions of parameters. On the other hand, fast training/testing and low computational cost are also necessary for quick deployment and development of COVID-19 screening systems, but traditional deep learning methods are usually computationally intensive. To address the above problems, we propose MiniSeg, a lightweight deep learning model for efficient COVID-19 segmentation. Compared with traditional segmentation methods, MiniSeg has several significant strengths: i) it only has 83K parameters and is thus not easy to overfit; ii) it has high computational efficiency and is thus convenient for practical deployment; iii) it can be fast retrained by other users using their private COVID-19 data for further improving performance. In addition, we build a comprehensive COVID-19 segmentation benchmark for comparing MiniSeg to traditional methods.

cs.CV↗

arXiv2021

Point-supervised Segmentation of Microscopy Images and Volumes via Objectness Regularization

Annotation is a major hurdle in the semantic segmentation of microscopy images and volumes due to its prerequisite expertise and effort. This work enables the training of semantic segmentation networks on images with only a single point for training per instance, an extreme case of weak supervision which drastically reduces the burden of annotation. Our approach has two key aspects: (1) we construct a graph-theoretic soft-segmentation using individual seeds to be used within a regularizer during training and (2) we use an objective function that enables learning from the constructed soft-labels. We achieve competitive results against the state-of-the-art in point-supervised semantic segmentation on challenging datasets in digital pathology. Finally, we scale our methodology to point-supervised segmentation in 3D fluorescence microscopy volumes, obviating the need for arduous manual volumetric delineation. Our code is freely available.

cs.CV↗

arXiv2021

Pose Refinement Graph Convolutional Network for Skeleton-based Action Recognition

With the advances in capturing 2D or 3D skeleton data, skeleton-based action recognition has received an increasing interest over the last years. As skeleton data is commonly represented by graphs, graph convolutional networks have been proposed for this task. While current graph convolutional networks accurately recognize actions, they are too expensive for robotics applications where limited computational resources are available. In this paper, we therefore propose a highly efficient graph convolutional network that addresses the limitations of previous works. This is achieved by a parallel structure that gradually fuses motion and spatial information and by reducing the temporal resolution as early as possible. Furthermore, we explicitly address the issue that human poses can contain errors. To this end, the network first refines the poses before they are further processed to recognize the action. We therefore call the network Pose Refinement Graph Convolutional Network. Compared to other graph convolutional networks, our network requires 86\%-93\% less parameters and reduces the floating point operations by 89%-96% while achieving a comparable accuracy. It therefore provides a much better trade-off between accuracy, memory footprint and processing time, which makes it suitable for robotics applications.

cs.CV↗

arXiv2020

MS-TCN++: Multi-Stage Temporal Convolutional Network for Action Segmentation

With the success of deep learning in classifying short trimmed videos, more attention has been focused on temporally segmenting and classifying activities in long untrimmed videos. State-of-the-art approaches for action segmentation utilize several layers of temporal convolution and temporal pooling. Despite the capabilities of these approaches in capturing temporal dependencies, their predictions suffer from over-segmentation errors. In this paper, we propose a multi-stage architecture for the temporal action segmentation task that overcomes the limitations of the previous approaches. The first stage generates an initial prediction that is refined by the next ones. In each stage we stack several layers of dilated temporal convolutions covering a large receptive field with few parameters. While this architecture already performs well, lower layers still suffer from a small receptive field. To address this limitation, we propose a dual dilated layer that combines both large and small receptive fields. We further decouple the design of the first stage from the refining stages to address the different requirements of these stages. Extensive evaluation shows the effectiveness of the proposed model in capturing long-range dependencies and recognizing action segments. Our models achieve state-of-the-art results on three datasets: 50Salads, Georgia Tech Egocentric Activities (GTEA), and the Breakfast dataset.

cs.CV↗

arXiv2019

Exploring Frame Segmentation Networks for Temporal Action Localization

Temporal action localization is an important task of computer vision. Though many methods have been proposed, it still remains an open question how to predict the temporal location of action segments precisely. Most state-of-the-art works train action classifiers on video segments pre-determined by action proposal. However, recent work found that a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. In this paper, we propose a Frame Segmentation Network (FSN) that places a temporal CNN on top of the 2D spatial CNNs. Spatial CNNs are responsible for abstracting semantics in spatial dimension while temporal CNN is responsible for introducing temporal context information and performing dense predictions. The proposed FSN can make dense predictions at frame-level for a video clip using both spatial and temporal context information. FSN is trained in an end-to-end manner, so the model can be optimized in spatial and temporal domain jointly. We also adapt FSN to use it in weakly supervised scenario (WFSN), where only video level labels are provided when training. Experiment results on public dataset show that FSN achieves superior performance in both frame-level action localization and temporal action localization.

cs.CV↗

arXiv2018

Mapping the Real Space Distributions of Galaxies in SDSS DR7: II. Measuring the growth rate, clustering amplitude of matter and biases of galaxies at redshift $0.1$

We extend the real-space mapping method developed in Shi et at. (2016) so that it can be applied to flux-limited galaxy samples. We use an ensemble of mock catalogs to demonstrate the reliability of this extension, showing that it allows for an accurate recovery of the real-space correlation functions and galaxy biases. We also demonstrate that, using an iterative method applied to intermediate-scale clustering data, we can obtain an unbiased estimate of the growth rate of structure $fσ_8$, which is related to the clustering amplitude of matter, to an accuracy of $\sim 10\%$. Applying this method to the Sloan Digital Sky Survey (SDSS) Data Release 7 (DR7), we construct a real-space galaxy catalog spanning the redshift range $0.01 \leq z \leq 0.2$, which contains 584,473 galaxies in the north Galactic cap (NGC). Using this data, we infer $\fss$ at a median redshift $z=0.1$, which is consistent with the WMAP9 cosmology at the $1σ$ level. By combining this measurement with the real-space clustering of galaxies and with galaxy-galaxy weak lensing measurements for the same sets of galaxies, we are able to break the degeneracy between $f$, $σ_8$, and $b$. From the SDSS DR7 data alone, we obtain the following cosmological constraints at redshift $z=0.1$: $f=$$0.464^{+0.040}_{-0.040}$, $σ_8=0.769^{+0.121}_{-0.089}$, and $b=1.910^{+0.234}_{-0.268}$, $1.449^{+0.194}_{-0.196}$, $1.301^{+0.170}_{-0.177}$, and $1.196^{+0.159}_{-0.161}~$ for galaxies within different absolute magnitude bins $^{0.1}{\rm M}_r-5\log h=[-23,0, -22.0], [-22,0, -21.0], [-21.0, -20.0]$ and $[-20.0, -19.0]$, respectively.

astro-ph.CO↗

arXiv2018

Full-sky ray-tracing simulation of weak lensing using ELUCID simulations: exploring galaxy intrinsic alignment and cosmic shear correlations

The intrinsic alignment of galaxies is an important systematic effect in weak-lensing surveys, which can affect the derived cosmological parameters. One direct way to distinguish different alignment models and quantify their effects on the measurement is to produce mocked weak-lensing surveys. In this work, we use full-sky ray-tracing technique to produce mock images of galaxies from the ELUCID $N$-body simulation run with the WMAP9 cosmology. In our model we assume that the shape of central elliptical galaxy follows that of the dark matter halo, and spiral galaxy follows the halo spin. Using the mocked galaxy images, a combination of galaxy intrinsic shape and the gravitational shear, we compare the predicted tomographic shear correlations to the results of KiDS and DLS. It is found that our predictions stay between the KiDS and DLS results. We rule out a model in which the satellite galaxies are radially aligned with the center galaxy, otherwise the shear-correlations on small scales are too high. Most important, we find that although the intrinsic alignment of spiral galaxies is very weak, they induce a positive correlation between the gravitational shear signal and the intrinsic galaxy orientation (GI). This is because the spiral galaxy is tangentially aligned with the nearby large-scale overdensity, contrary to the radial alignment of elliptical galaxy. Our results explain the origin of detected positive GI term from the weak-lensing surveys. We conclude that in future analysis, the GI model must include the dependence on galaxy types in more detail.

astro-ph.CO↗

arXiv2018

ELUCID V. Lighting dark matter halos with galaxies

In a recent study, using the distribution of galaxies in the north galactic pole of SDSS DR7 region enclosed in a 500$\mpch$ box, we carried out our ELUCID simulation (Wang et al. 2016, ELUCID III). Here we {\it light} the dark matter halos and subhalos in the reconstructed region in the simulation with galaxies in the SDSS observations using a novel {\it neighborhood} abundance matching method. Before we make use of thus established galaxy-subhalo connections in the ELUCID simulation to evaluate galaxy formation models, we set out to explore the reliability of such a link. For this purpose, we focus on the following a few aspects of galaxies: (1) the central-subhalo luminosity and mass relations; (2) the satellite fraction of galaxies; (3) the conditional luminosity function (CLF) and conditional stellar mass function (CSMF) of galaxies; and (4) the cross correlation functions between galaxies and the dark matter particles, most of which are measured separately for all, red and blue galaxy populations. We find that our neighborhood abundance matching method accurately reproduces the central-subhalo relations, satellite fraction, the CLFs and CSMFs and the biases of galaxies. These features ensure that thus established galaxy-subhalo connections will be very useful in constraining galaxy formation processes. And we provide some suggestions on the three levels of using the galaxy-subhalo pairs for galaxy formation constraints. The galaxy-subhalo links and the subhalo merger trees in the SDSS DR7 region extracted from our ELUCID simulation are available upon request.

astro-ph.GA↗

arXiv2017

ELUCID IV: Galaxy Quenching and its Relation to Halo Mass, Environment, and Assembly Bias

We examine the quenched fraction of central and satellite galaxies as a function of galaxy stellar mass, halo mass, and the matter density of their large scale environment. Matter densities are inferred from our ELUCID simulation, a constrained simulation of local Universe sampled by SDSS, while halo masses and central/satellite classification are taken from the galaxy group catalog of Yang et al. The quenched fraction for the total population increases systematically with the three quantities. We find that the `environmental quenching efficiency', which quantifies the quenched fraction as function of halo mass, is independent of stellar mass. And this independence is the origin of the stellar mass-independence of density-based quenching efficiency, found in previous studies. Considering centrals and satellites separately, we find that the two populations follow similar correlations of quenching efficiency with halo mass and stellar mass, suggesting that they have experienced similar quenching processes in their host halo. We demonstrate that satellite quenching alone cannot account for the environmental quenching efficiency of the total galaxy population and the difference between the two populations found previously mainly arises from the fact that centrals and satellites of the same stellar mass reside, on average, in halos of different mass. After removing these halo-mass and stellar-mass effects, there remains a weak, but significant, residual dependence on environmental density, which is eliminated when halo assembly bias is taken into account. Our results therefore indicate that halo mass is the prime environmental parameter that regulates the quenching of both centrals and satellites.

astro-ph.GA↗

arXiv2017

Revealing the cosmic web dependent halo bias

Halo bias is the one of the key ingredients of the halo models. It was shown at a given redshift to be only dependent, to the first order, on the halo mass. In this study, four types of cosmic web environments: clusters, filaments, sheets and voids are defined within a state of the art high resolution $N$-body simulation. Within those environments, we use both halo-dark matter cross-correlation and halo-halo auto correlation functions to probe the clustering properties of halos. The nature of the halo bias differs strongly among the four different cosmic web environments we describe. With respect to the overall population, halos in clusters have significantly lower biases in the {$10^{11.0}\sim 10^{13.5}h^{-1}\rm M_\odot$} mass range. In other environments however, halos show extremely enhanced biases up to a factor 10 in voids for halos of mass {$\sim 10^{12.0}h^{-1}\rm M_\odot$}. Such a strong cosmic web environment dependence in the halo bias may play an important role in future cosmological and galaxy formation studies. Within this cosmic web framework, the age dependency of halo bias is found to be only significant in clusters and filaments for relatively small halos $\la 10^{12.5}\msunh$.

astro-ph.GA↗

arXiv2017

ELUCID -- Exploring the Local Universe with the reConstructed Initial Density field. II: Reconstruction diagnostics, applied to numerical halo catalogs

The ELUCID project aims to build a series of realistic cosmological simulations that reproduce the spatial and mass distribution of the galaxies as observed in the Sloan Digital Sky Survey (SDSS). This requires powerful reconstruction techniques to create constrained initial conditions. We test the reconstruction method by applying it to several $N$-body simulations. We use 2 medium resolution simulations from each of which three additional constrained $N$-body simulations were produced. We compare the resulting friend of friend catalogs by using the particle indexes as tracers, and quantify the quality of the reconstruction by varying the main smoothing parameter. The cross identification method we use proves to be efficient, and the results suggest that the most massive reconstructed halos are effectively traced from the same Lagrangian regions in the initial conditions. Preliminary time dependence analysis indicates that high mass end halos converge only at a redshift close to the reconstruction redshift. This suggests that, for earlier snapshots, only collections of progenitors may be effectively cross-identified.

astro-ph.CO↗