Pain Intensity Estimation Research Articles

Pain assessment from facial expression images utilizing Statistical Frei-Chen Mask (SFCM)-based features and DenseNet

Estimating pain levels is crucial for patients with serious illnesses, those recovering from brain surgery, and those receiving intensive care etc. An automatic pain intensity estimator is proposed in this study that gathers information about pain and intensity from the user’s expressions. The faces in the database are first cropped using a ‘Chehra’ face detector, which performs well even in wildly uncontrolled environments with a wide range of lighting and position fluctuations. The suggested technique extracts the beneficial and distinct patterns from facial expressions using novel Statistical Frei-Chen Mask (SFCM)-based features and DenseNet-based features. As it offers quick as well as accurate pain identification and pain intensity estimation, the Radial Basis Function Based Extreme Learning Machine (RBF-ELM) is employed for pain recognition and pain intensity level estimation using the characteristics. All the data is kept, updated and protected in the cloud because availability and high-performance decision-making are so important for informing physicians and auxiliary IoT nodes (such as wearable sensors). In addition, cloud computing reduces the time complexity of the training phase of Machine Learning algorithms in situations where it is possible to build a complete cloud/edge architecture by allocating additional computational resources and memory in use. The facial expression images from the UNBC-McMaster Shoulder Pain Expression Archive and 2D face dataset are used to test the proposed method. The measurement of pain intensity uses four stages. When compared to the results from the literature, the proposed work attains enhanced performance.

Global-local combined features to detect pain intensity from facial expression images with attention mechanism

The estimation of pain intensity is critical for medical diagnosis and treatment of patients. With the development of image monitoring technology and artificial intelligence, automatic pain assessment based on facial expression and behavioral analysis shows a potential value in clinical applications. This paper reports a framework of convolutional neural network with global and local attention mechanism (GLA-CNN) for the effective detection of pain intensity at four-level thresholds using facial expression images. GLA-CNN includes two modules, namely global attention network (GANet) and local attention network (LANet). LANet is responsible for extracting representative local patch features of faces, while GANet extracts whole facial features to compensate for the ignored correlative features between patches. In the end, the global correlational and local subtle features are fused for the final estimation of pain intensity. Experiments under the UNBC-McMaster Shoulder Pain database demonstrate that GLA-CNN outperforms other state-of-the-art methods. Additionally, a visualization analysis is conducted to present the feature map of GLA-CNN, intuitively showing that it can extract not only local pain features but also global correlative facial ones. Our study demonstrates that pain assessment based on facial expression is a non-invasive and feasible method, and can be employed as an auxiliary pain assessment tool in clinical practice.

Uncertainty quantification in neural-network based pain intensity estimation.

Improper pain management leads to severe physical or mental consequences, including suffering, a negative impact on quality of life, and an increased risk of opioid dependency. Assessing the presence and severity of pain is imperative to prevent such outcomes and determine the appropriate intervention. However, the evaluation of pain intensity is a challenging task because different individuals experience pain differently. To overcome this, many researchers in the field have employed machine learning models to evaluate pain intensity objectively using physiological signals. However, these efforts have primarily focused on pain point estimation, disregarding inherent uncertainty and variability in the data and model. A point estimate, which provides only partial information, is not sufficient for sound clinical decision-making. This study proposes a neural network-based method for objective pain interval estimation, and quantification of uncertainty. Our approach, which enables objective pain intensity estimation with desired confidence probabilities, affords clinicians a better understanding of a person's pain intensity. We explored three distinct algorithms: the bootstrap method, lower and upper bound estimation (LossL) optimized by genetic algorithm, and modified lower and upper bound estimation (LossS) optimized by gradient descent algorithm. Our empirical results demonstrate that LossS outperforms the other two by providing narrower prediction intervals. For 50%, 75%, 85%, and 95% prediction interval coverage probability, LossS provides average interval widths that are 22.4%, 7.9%, 16.7%, and 9.1% narrower than those of LossL, and 19.3%, 21.1%, 23.6%, and 26.9% narrower than those of bootstrap. As LossS outperforms, we assessed its performance in three different model-building approaches: (1) a generalized approach using a single model for the entire population, (2) a personalized approach with separate models for each individual, and (3) a hybrid approach with models for clusters of individuals. Results demonstrate that the hybrid model-building approach provides the best performance.

Does a diagnosis of depression influence observer ratings of pain severity? The mediating role of causal attributions of pain and pain genuineness.

Researchers have been increasingly investigating observer and patient characteristics that may influence the assessment of pain in others. While rates of psychiatric conditions are high in chronic pain populations, surprisingly little attention has been given to if (and why) a comorbid psychiatric diagnosis may influence the estimation of pain in others. Using an experimental vignette paradigm, the current study examined whether a diagnostic label of major depressive disorder (MDD) would impact observer pain estimates of a woman with chronic pain, and whether causal attributions of pain and pain genuineness might help explain these effects. Participants (n = 188) were given a vignette describing a female patient with chronic pain (who either had MDD or no mental health concerns), viewed a brief video clip of the patient, and then were asked to provide a variety of ratings about the woman's pain. Results of a serial multiple mediation analysis revealed that participants in the MDD condition made greater psychological attributions for the woman's pain, which was associated with lower perceptions of pain genuineness, which was then associated with lower estimates of pain intensity. These findings suggest that a diagnosis of depression may indirectly influence observer estimates of another person's pain by heightening psychological attributions of pain, and making their pain seem less genuine. Further research is needed to elucidate the complex processes underlying pain estimation, including patient and observer characteristics, biases, and heuristics, in order to improve quality of care for those living with persistent pain.'

Use of Wearable Sensor Device and Mobile Application for Objective Assessment of Pain in Post-surgical Patients: A Preliminary Study.

Effective post-operative pain management requires an accurate and frequent assessment of the pain experienced by the patients. The current gold-standard of pain assessment is through patient self-evaluation (e.g., numeric rating scale, NRS) which is subjective, prone to recall-bias, and does not provide comprehensive information of the pain intensity and its trends. We conducted a study to explore the potential of wearable biosensors and machine learning-based analysis of physiological parameters to estimate the pain intensity. The results from our study of post-operative knee surgery patients monitored over a period of 30 days demonstrate the feasibility of the system in ambulatory setting, with a substantial agreement (Cohen's Kappa = 0.70, 95% CI 0.68-0.72) between the pain intensity estimation and the patient reported numerical rating scale. Therefore, the wearable biosensors coupled with the machine learning-derived pain estimation are capable of remotely assessing the pain intensity.

Current attitudes and self-rated abilities of Bosnia and Herzegovina veterinarians toward pain recognition and quantification in domestic animals.

In previous years interest has grown in investigating the attitudes and capabilities of veterinarians regarding the recognition, quantification and treatment of animal pain throughout different parts of the world and encompassing various species. This is the first report exploring the attitudes and self-rated abilities of veterinarians in Bosnia and Herzegovina (B&H) concerning recognition and quantification of pain in domestic animals. A study questionnaire was made available to 535 general practice veterinarians throughout B&H and 73 (14%) responded in full. The questionnaire contained polar, multiple choice, ordinal and interval scale questions and consisted of sections asking about demographic data, attitudes to pain recognition and quantification, use and availability of analgesics, estimates of pain intensity during specific surgical procedures, and the perceived need for pain assessment and continuing education programmes for analgesia. Half of the respondents considered the recognition and quantification of pain to be difficult while 89% did not make use of pain assessment scales. Of the respondents, (33/73; 45%) felt a certain level of pain to be advantageous since it reduces the activity of the healing animal, whereas 52% (38/73) did not agreed with this concept. Cost was a consideration when deciding whether or not to use analgesics for 58% (42/73) of the respondents with the most commonly used types being NSAIDs (72/73;99%) and opioids (60/73; 82%). Practitioners in B&H displayed awareness of the importance of pain assessment and management however a significant proportion were unaware of pain scales and relied upon physiological indicators of pain.

Image-Based Pain Intensity Estimation Using Parallel CNNs with Regional Attention.

Automatic pain estimation plays an important role in the field of medicine and health. In the previous studies, most of the entire image frame was directly imported into the model. This operation can allow background differences to negatively affect the experimental results. To tackle this issue, we propose the parallel CNNs framework with regional attention for automatic pain intensity estimation at the frame level. This modified convolution neural network structure combines BlurPool methods to enhance translation invariance in network learning. The improved networks can focus on learning core regions while supplementing global information, thereby obtaining parallel feature information. The core regions are mainly based on the tradeoff between the weights of the channel attention modules and the spatial attention modules. Meanwhile, the background information of the non-core regions is shielded by the DropBlock algorithm. These steps enable the model to learn facial pain features adaptively, not limited to a single image pattern. The experimental result of our proposed model outperforms many state-of-the-art methods on the RMSE and PCC metrics when evaluated on the diverse pain levels of over 12,000 images provided by the publicly available UNBC dataset. The model accuracy rate has reached 95.11%. The experimental results show that the proposed method is highly efficient at extracting the facial features of pain and predicts pain levels with high accuracy.

Smoothed Generalized Dirichlet: A Novel Count-Data Model for Detecting Emotional States

In this article, we propose novel approaches to deal with the problem of burstiness, the challenge of count-data sparseness, and the curse of dimensionality. We introduce a smoothed generalized Dirichlet distribution that is a smoothed variant of the generalized Dirichlet distribution and a generalization of the smoothed Dirichlet. We provide different learning methods based on mixture models and agglomerative clustering-based geometrical information: Kullback–Leibler divergence, Fisher metric, and Bhattacharyya distance. Moreover, we show that the new smoothed generalized Dirichlet could be considered as a prior to the multinomial, which generates a new distribution for count data that we call the smoothed generalized Dirichlet multinomial. In particular, we present an approximation based on Taylor series expansion for better performance and optimized running time in the case of high-dimensional count data. The proposed models are evaluated through two emotion detection applications: disaster-tweet-related emotions and pain intensity estimation. Experiments show the efficiency and the robustness of our approaches when dealing with texts, videos, and images.

Convolution Neural Network for Pain Intensity Assessment from Facial Expression.

Pain is an unpleasant feeling that can reflect a patient's health situation. Since measuring pain is subjective, time-consuming, and needs continuous monitoring, automated pain intensity detection from facial expression holds great potential for smart healthcare applications. Convolutional Neural Networks (CNNs) are recently being used to identify features, map and model pain intensity from facial images, delivering great promise in helping practitioners detect disease. Limited research has been conducted to determine pain intensity levels across multiple classes. CNNs with simple learning schemes are limited in their ability to extract feature information from images. In order to develop a highly accurate pain intensity estimation system, this study proposes a Deep CNN (DCNN) model using the transfer learning technique, where a pre-trained DCNN model is adopted by replacing its dense upper layers, and the model is tuned using painful facial. We conducted experiments on the UNBC-McMaster shoulder pain archive database to estimate pain intensity in terms of seven-level thresholds using a given facial expression image. The experiments show our method achieves a promising improvement in terms of accuracy and performance to estimate pain intensity and outperform the-state-of-the-arts models.

Experimental Exploration of Objective Human Pain Assessment Using Multimodal Sensing Signals.

Optimization of pain assessment and treatment is an active area of research in healthcare. The purpose of this research is to create an objective pain intensity estimation system based on multimodal sensing signals through experimental studies. Twenty eight healthy subjects were recruited at Northeastern University. Nine physiological modalities were utilized in this research, namely facial expressions (FE), electroencephalography (EEG), eye movement (EM), skin conductance (SC), and blood volume pulse (BVP), electromyography (EMG), respiration rate (RR), skin temperature (ST), blood pressure (BP). Statistical analysis and machine learning algorithms were deployed to analyze the physiological data. FE, EEG, SC, BVP, and BP proved to be able to detect different pain states from healthy subjects. Multi-modalities proved to be promising in detecting different levels of painful states. A decision-level multi-modal fusion also proved to be efficient and accurate in classifying painful states.

Exploring Parental Responses to Pre-schoolers' "Everyday" Pain Experiences Through Electronic Diary and Ecological Momentary Assessment Methodologies.

Objective: Parental influence during children’s “everyday” pain events is under-explored, compared to clinical or experimental pains. We trialed two digital reporting methods for parents to record the real-world context surrounding their child’s everyday pain events within the family home.Methods: Parents (N = 21) completed a structured e-diary for 14 days, reporting on one pain event experienced by their child (aged 2.5–6 years) each day, and describing child pain responses, parental supervision, parental estimates of pain severity and intensity, and parental catastrophizing, distress, and behavioral responses. During the same 2-week period, a subsample of parent-child pairs (N = 9) completed digital ecological momentary assessments (EMA), immediately after any chosen pain event. Children reported their current pain while parents estimated the child’s pain and indicated their own distress.Results: “Everyday” pain events frequently featured minor injuries to the child’s head, hands or knees, and child responses included crying and non-verbal comments (e.g., “Ouch!”). Pain events occurred less frequently when parents had been supervising their child, and supervising parents reported lower levels of worry and anxiety than non-supervising parents. Child sex was significantly associated with parental estimates of pain intensity, with parents of girls giving higher estimates than parents of boys. Child age was significantly associated with both the number of pain events and with parental estimates of pain intensity and child distress: the youngest children (2–3 years) experienced the fewest pain events but received higher pain and distress estimates from parents than older children. Hierarchal Linear Modeling revealed that parental estimates of pain severity were significant positive predictors of parental distress and catastrophizing in response to a specific pain event. Furthermore, higher levels of parental catastrophic thinking in response to a specific pain event resulted in increased distress, solicitousness, and coping-promoting behaviors in parents. The EMA data revealed that children reported significantly higher pain intensity than their parents.Conclusion: The electronic pain diary provided a key insight into the nature of “everyday” pain experiences around the family home. Digital daily reporting of how the family copes with “everyday” events represents a viable means to explore a child’s everyday pains without disrupting their home environment.

Computer aided pain detection and intensity estimation using compact CNN based fusion network

Pain is an indication of uneasiness, and its assessment is crucial for medical diagnosis and treatment of patients. Subjective methods like visual inspection and self-report of pain, are prone to human errors. Hence, objective methods i.e. behavioural, physiological are being focused on by researchers in the recent past. In this paper, a behavioural indicator i.e. facial expression-based fully automated pain assessment system is proposed. This system includes a novel fusion structure of Convolutional Networks for assessing various pain intensities from raw facial images. Here, we have proposed joint learning of robust pain-related facial expression features from fused RGB appearance and shape-based latent representation. We have suggested that the joint learning from both feature representations results in a more robust pain assessment model as compared to learning from either of the representation independently. Given this, the proposed system used two shallower networks, i.e., Spatial Appearance Network and Shape Descriptor Network. The proposed model has been evaluated extensively on the UNBC-McMaster dataset for both pain classification and intensity estimation. For pain level classification the proposed technique achieved 0.94 of F1-score. Moreover, for pain intensity estimation the proposed model has achieved a mean absolute error of 0.22 and accuracy of 0.92.

Exploration of physiological sensors, features, and machine learning models for pain intensity estimation.

In current clinical settings, typically pain is measured by a patient’s self-reported information. This subjective pain assessment results in suboptimal treatment plans, over-prescription of opioids, and drug-seeking behavior among patients. In the present study, we explored automatic objective pain intensity estimation machine learning models using inputs from physiological sensors. This study uses BioVid Heat Pain Dataset. We extracted features from Electrodermal Activity (EDA), Electrocardiogram (ECG), Electromyogram (EMG) signals collected from study participants subjected to heat pain. We built different machine learning models, including Linear Regression, Support Vector Regression (SVR), Neural Networks and Extreme Gradient Boosting for continuous value pain intensity estimation. Then we identified the physiological sensor, feature set and machine learning model that give the best predictive performance. We found that EDA is the most information-rich sensor for continuous pain intensity prediction. A set of only 3 features from EDA signals using SVR model gave an average performance of 0.93 mean absolute error (MAE) and 1.16 root means square error (RMSE) for the subject-independent model and of 0.92 MAE and 1.13 RMSE for subject-dependent. The MAE achieved with signal-feature-model combination is less than 1 unit on 0 to 4 continues pain scale, which is smaller than the MAE achieved by the methods reported in the literature. These results demonstrate that it is possible to estimate pain intensity of a patient using a computationally inexpensive machine learning model with 3 statistical features from EDA signal which can be collected from a wrist biosensor. This method paves a way to developing a wearable pain measurement device.

Deep domain adaptation with ordinal regression for pain assessment using weakly-labeled videos

Estimation of pain intensity from facial expressions captured in videos has an immense potential for health care applications. Given the challenges related to subjective variations of facial expressions, and to operational capture conditions, the accuracy of state-of-the-art deep learning (DL) models for recognizing facial expressions may decline. Domain adaptation (DA) has been widely explored to alleviate the problem of domain shifts that typically occur between video data captured across various source (laboratory) and target (operational) domains. Moreover, given the laborious task of collecting and annotating videos, and the subjective bias due to ambiguity among adjacent intensity levels, weakly-supervised learning (WSL) is gaining attention in such applications. State-of-the-art WSL models are typically formulated as regression problems, and do not leverage the ordinal relationship among pain intensity levels, nor the temporal coherence of multiple consecutive frames. This paper introduces a new DL model for weakly-supervised DA with ordinal regression (WSDA-OR) that can be adapted using target domain videos with coarse labels provided on a periodic basis. The WSDA-OR model enforces ordinal relationships among the intensity levels assigned to target sequences, and associates multiple relevant frames to sequence-level labels (instead of a single frame). In particular, it learns discriminant and domain-invariant feature representations by integrating multiple instance learning with deep adversarial DA, where soft Gaussian labels are used to efficiently represent the weak ordinal sequence-level labels from the target domain. The proposed approach was validated using the RECOLA video dataset as fully-labeled source domain data, and UNBC-McMaster shoulder pain video dataset as weakly-labeled target domain data. We have also validated WSDA-OR on BIOVID and Fatigue (private) datasets for sequence level estimation. Experimental results indicate that our proposed approach can significantly improve performance over the state-of-the-art models, allowing to achieve a greater pain localization accuracy. Code is available on GitHub link: https://github.com/praveena2j/WSDAOR.

HybNet: a hybrid network structure for pain intensity estimation

Automatic pain intensity estimation has great potential in current rehabilitation medicine, and patients’ health status information can be obtained through the analysis of facial images. At present, deep convolutional neural networks (CNNs) have made great progress in many fields, including natural language processing, image classification and action recognition. Motivated by the current achievements, a novel end-to-end hybrid network is proposed to extract multidimensional features from image sequences, which is composed of 3D convolution, 2D convolution and 1D convolution. Specifically, the 3D convolutional neural network (3D CNN) is designed to capture the spatiotemporal features, and the 2D convolutional neural network (2D CNN) is designed to capture the spatial features, while the 1D convolutional neural network (1D CNN) is mainly used to capture the geometric information from facial landmarks. Finally, the features obtained by the three different networks are fused together for regression. The proposed HybNet is evaluated on UNBC-McMaster Shoulder Pain Expression Archive Database, and the experimental results show that it can effectively extract the discriminative high-level features and can achieve competitive performance with the state-of-the-art methods.

PLAAN: Pain Level Assessment with Anomaly-detection based Network

Automatic chronic pain assessment and pain intensity estimation has been attracting growing attention due to its widespread applications. One of the prevalent issues in automatic pain analysis is inadequate balanced expert-labelled data for pain estimation. This work proposes an anomaly detection based network addressing one of the existing limitations of automatic pain assessment. The evaluation of the network is performed on pain intensity estimation and protective behaviour estimation tasks from body movements in the EmoPain Challenge dataset. The EmoPain dataset consists of body part based sensor data for both the tasks. The proposed network, PLAAN (Pain Level Assessment with Anomaly-detection based Network), is a lightweight LSTM-DNN network which considers features based on sensor data as the input and predicts intensity level of pain and presence or absence of protective behaviour in chronic low back pain patients. Joint training considering body movement patterns, such as exercise type, corresponding to pain exhibition as a label improves the performance of the network. However, contrary to perception, protective behaviour rather exists sporadically alongside pain in the EmoPain dataset. This induces yet another complication in accurate estimation of protective behaviour. This problem is resolved by incorporating anomaly detection in the network. A detailed comparison of different networks with varied features is outlined in the paper, presenting a significant improvement with the final proposed anomaly detection based network.

A Personalized Spatial-Temporal Cold Pain Intensity Estimation Model Based on Facial Expression.

Objective: Pain assessment is of great importance in both clinical research and patient care. Facial expression analysis is becoming a key part of pain detection because it is convenient, automatic, and real-time. The aim of this study is to present a cold pain intensity estimation experiment, investigate the importance of the spatial-temporal information on facial expression based cold pain, and study the performance of the personalized model as well as the generalized model. Methods: A cold pain experiment was carried out and facial expressions from 29 subjects were extracted. Three different architectures (Inception V3, VGG-LSTM, and Convolutional LSTM) were used to estimate three intensities of cold pain: No pain, Moderate pain, and Severe Pain. Architectures with Sequential information were compared with single-frame architecture, showing the importance of spatial-temporal information on pain estimation. The performances of the personalized model and the generalized model were also compared. Results: A mean F1 score of 79.48% was achieved using Convolutional LSTM based on the personalized model. Conclusion: This study demonstrates the potential for the estimation of cold pain intensity from facial expression analysis and shows that the personalized spatial-temporal framework has better performance in cold pain intensity estimation. Significance: This cold pain intensity estimator could allow convenient, automatic, and real-time use to provide continuous objective pain intensity estimations of subjects and patients.

The modeling of human facial pain intensity based on Temporal Convolutional Networks trained with video frames in HSV color space

An accurate detection and management of pain, measured through its relative intensity, plays an important role in the treatment of disease and reducing a patient’s discomfort. As it is relatively difficult to assess, describe, evaluate and manage the pain level using a patient’s self-report, automated pain-detecting tools can provide useful information to assist in the management of pain intensity. This study proposes a new predictive modeling framework that employs a modified Temporal Convolutional Network (TCN) algorithm to recognize the pain intensity prevalent in patients’ video frames collected as part of UNBC-McMaster Shoulder Pain Archive and MIntPAIN databases. The inputs of the proposed TCN network is composed of the extracted and reduced face image features from a fine-tuned VGG-Face and principal component analysis (PCA) with Hue, Saturation, Value (HSV) color spaces video images. The results of TCN based predictive model, employing a long short-term memory (LSTM) model as well as other state-of-the art models, show that the proposed approach performs faster with a high level of efficiency. This is demonstrated by the low magnitude of error metrics (i.e., Mean Squared Error = 0.0629, Mean Absolute Error = 0.1021, correctness validation results represented by Area under Curve = 85% and accuracy metric = 92.44%). Considering the efficiency of the proposed TCN framework, integrating fine-tuned VGG-Face and PCA with Hue, Saturation, Value (HSV) color spaces video images for pain intensity estimation, the present study affirms that the new method can be adopted as an automatic health informatics tool, mainly for pain detection, and subsequently, implemented in the pain management area.

Self-supervised pain intensity estimation from facial videos via statistical spatiotemporal distillation

• We propose a self-supervised representation learning for pain intensity estimation from the face. • We use a novel similarity function in a Siamese network to learn representations from unlabeled data as the pretext task. • Statistical Spatiotemporal Distillation is presented to encode dynamic and appearance information of video into an image map. • Our proposed method shows promising performance in cross-dataset evaluation setting using two publicly available datasets. Recently, automatic pain assessment technology, in particular automatically detecting pain from facial expressions, has been developed to improve the quality of pain management, and has attracted increasing attention. In this paper, we propose self-supervised learning for automatic yet efficient pain assessment, in order to reduce the cost of collecting large amount of labeled data. To achieve this, we introduce a novel similarity function to learn generalized representations using a Siamese network in the pretext task. The learned representations are finetuned in the downstream task of pain intensity estimation. To make the method computationally efficient, we propose Statistical Spatiotemporal Distillation (SSD) to encode the spatiotemporal variations underlying the facial video into a single RGB image, enabling the use of less complex 2D deep models for video representation. Experiments on two publicly available pain datasets and cross-dataset evaluation demonstrate promising results, showing the good generalization ability of the learned representations.

Pain intensity estimation based on a spatial transformation and attention CNN.

Models designed to detect abnormalities that reflect disease from facial structures are an emerging area of research for automated facial analysis, which has important potential value in smart healthcare applications. However, most of the proposed models directly analyze the whole face image containing the background information, and rarely consider the effects of the background and different face regions on the analysis results. Therefore, in view of these effects, we propose an end-to-end attention network with spatial transformation to estimate different pain intensities. In the proposed method, the face image is first provided as input to a spatial transformation network for solving the problem of background interference; then, the attention mechanism is used to adaptively adjust the weights of different face regions of the transformed face image; finally, a convolutional neural network (CNN) containing a Softmax function is utilized to classify the pain levels. The extensive experiments and analysis are conducted on the benchmarking and publicly available database, namely the UNBC-McMaster shoulder pain. More specifically, in order to verify the superiority of our proposed method, the comparisons with the basic CNNs and the-state-of-the-arts are performed, respectively. The experiments show that the introduced spatial transformation and attention mechanism in our method can significantly improve the estimation performances and outperform the-state-of-the-arts.