Searcharxiv⌕ Search

arXiv subjects

Reza Azad

Publications and source records attributed to Reza Azad.

54 records · Page 3Linked to original sources
arXiv2021

Stacked Hourglass Network with a Multi-level Attention Mechanism: Where to Look for Intervertebral Disc Labeling

Labeling vertebral discs from MRI scans is important for the proper diagnosis of spinal related diseases, including multiple sclerosis, amyotrophic lateral sclerosis, degenerative cervical myelopathy and cancer. Automatic labeling of the vertebral discs in MRI data is a difficult task because of the similarity between discs and bone area, the variability in the geometry of the spine and surrounding tissues across individuals, and the variability across scans (manufacturers, pulse sequence, image contrast, resolution and artefacts). In previous studies, vertebral disc labeling is often done after a disc detection step and mostly fails when the localization algorithm misses discs or has false positive detection. In this work, we aim to mitigate this problem by reformulating the semantic vertebral disc labeling using the pose estimation technique. To do so, we propose a stacked hourglass network with multi-level attention mechanism to jointly learn intervertebral disc position and their skeleton structure. The proposed deep learning model takes into account the strength of semantic segmentation and pose estimation technique to handle the missing area and false positive detection. To further improve the performance of the proposed method, we propose a skeleton-based search space to reduce false positive detection. The proposed method evaluated on spine generic public multi-center dataset and demonstrated better performance comparing to previous work, on both T1w and T2w contrasts. The method is implemented in ivadomed (https://ivadomed.org).

eess.IV↗
arXiv2021

Multi-scale Regional Attention Deeplab3+: Multiple Myeloma Plasma Cells Segmentation in Microscopic Images

Multiple myeloma cancer is a type of blood cancer that happens when the growth of abnormal plasma cells becomes out of control in the bone marrow. There are various ways to diagnose multiple myeloma in bone marrow such as complete blood count test (CBC) or counting myeloma plasma cell in aspirate slide images using manual visualization or through image processing technique. In this work, an automatic deep learning method for the detection and segmentation of multiple myeloma plasma cell have been explored. To this end, a two-stage deep learning method is designed. In the first stage, the nucleus detection network is utilized to extract each instance of a cell of interest. The extracted instance is then fed to the multi-scale function to generate a multi-scale representation. The objective of the multi-scale function is to capture the shape variation and reduce the effect of object scale on the cytoplasm segmentation network. The generated scales are then fed into a pyramid of cytoplasm networks to learn the segmentation map in various scales. On top of the cytoplasm segmentation network, we included a scale aggregation function to refine and generate a final prediction. The proposed approach has been evaluated on the SegPC2021 grand-challenge and ranked second on the final test phase among all teams.

eess.IV↗
arXiv2020

On the Texture Bias for Few-Shot CNN Segmentation

Despite the initial belief that Convolutional Neural Networks (CNNs) are driven by shapes to perform visual recognition tasks, recent evidence suggests that texture bias in CNNs provides higher performing models when learning on large labeled training datasets. This contrasts with the perceptual bias in the human visual cortex, which has a stronger preference towards shape components. Perceptual differences may explain why CNNs achieve human-level performance when large labeled datasets are available, but their performance significantly degrades in lowlabeled data scenarios, such as few-shot semantic segmentation. To remove the texture bias in the context of few-shot learning, we propose a novel architecture that integrates a set of Difference of Gaussians (DoG) to attenuate high-frequency local components in the feature space. This produces a set of modified feature maps, whose high-frequency components are diminished at different standard deviation values of the Gaussian distribution in the spatial domain. As this results in multiple feature maps for a single image, we employ a bi-directional convolutional long-short-term-memory to efficiently merge the multi scale-space representations. We perform extensive experiments on three well-known few-shot segmentation benchmarks -- Pascal i5, COCO-20i and FSS-1000 -- and demonstrate that our method outperforms state-of-the-art approaches in two datasets under the same conditions. The code is available at: https://github.com/rezazad68/fewshot-segmentation

cs.CV↗
arXiv2020

Semi-supervised few-shot learning for medical image segmentation

Recent years have witnessed the great progress of deep neural networks on semantic segmentation, particularly in medical imaging. Nevertheless, training high-performing models require large amounts of pixel-level ground truth masks, which can be prohibitive to obtain in the medical domain. Furthermore, training such models in a low-data regime highly increases the risk of overfitting. Recent attempts to alleviate the need for large annotated datasets have developed training strategies under the few-shot learning paradigm, which addresses this shortcoming by learning a novel class from only a few labeled examples. In this context, a segmentation model is trained on episodes, which represent different segmentation problems, each of them trained with a very small labeled dataset. In this work, we propose a novel few-shot learning framework for semantic segmentation, where unlabeled images are also made available at each episode. To handle this new learning paradigm, we propose to include surrogate tasks that can leverage very powerful supervisory signals --derived from the data itself-- for semantic feature learning. We show that including unlabeled surrogate tasks in the episodic training leads to more powerful feature representations, which ultimately results in better generability to unseen tasks. We demonstrate the efficiency of our method in the task of skin lesion segmentation in two publicly available datasets. Furthermore, our approach is general and model-agnostic, which can be combined with different deep architectures.

cs.CV↗
arXiv2020

Multi-level Context Gating of Embedded Collective Knowledge for Medical Image Segmentation

Medical image segmentation has been very challenging due to the large variation of anatomy across different cases. Recent advances in deep learning frameworks have exhibited faster and more accurate performance in image segmentation. Among the existing networks, U-Net has been successfully applied on medical image segmentation. In this paper, we propose an extension of U-Net for medical image segmentation, in which we take full advantages of U-Net, Squeeze and Excitation (SE) block, bi-directional ConvLSTM (BConvLSTM), and the mechanism of dense convolutions. (I) We improve the segmentation performance by utilizing SE modules within the U-Net, with a minor effect on model complexity. These blocks adaptively recalibrate the channel-wise feature responses by utilizing a self-gating mechanism of the global information embedding of the feature maps. (II) To strengthen feature propagation and encourage feature reuse, we use densely connected convolutions in the last convolutional layer of the encoding path. (III) Instead of a simple concatenation in the skip connection of U-Net, we employ BConvLSTM in all levels of the network to combine the feature maps extracted from the corresponding encoding path and the previous decoding up-convolutional layer in a non-linear way. The proposed model is evaluated on six datasets DRIVE, ISIC 2017 and 2018, lung segmentation, $PH^2$, and cell nuclei segmentation, achieving state-of-the-art performance.

eess.IV↗
arXiv2019

Bi-Directional ConvLSTM U-Net with Densley Connected Convolutions

In recent years, deep learning-based networks have achieved state-of-the-art performance in medical image segmentation. Among the existing networks, U-Net has been successfully applied on medical image segmentation. In this paper, we propose an extension of U-Net, Bi-directional ConvLSTM U-Net with Densely connected convolutions (BCDU-Net), for medical image segmentation, in which we take full advantages of U-Net, bi-directional ConvLSTM (BConvLSTM) and the mechanism of dense convolutions. Instead of a simple concatenation in the skip connection of U-Net, we employ BConvLSTM to combine the feature maps extracted from the corresponding encoding path and the previous decoding up-convolutional layer in a non-linear way. To strengthen feature propagation and encourage feature reuse, we use densely connected convolutions in the last convolutional layer of the encoding path. Finally, we can accelerate the convergence speed of the proposed network by employing batch normalization (BN). The proposed model is evaluated on three datasets of: retinal blood vessel segmentation, skin lesion segmentation, and lung nodule segmentation, achieving state-of-the-art performance.

eess.IV↗
arXiv2014

A Robust and Efficient Method for Improving Accuracy of License Plate Characters Recognition

License Plate Recognition (LPR) plays an important role on the traffic monitoring and parking management. A robust and efficient method for enhancing accuracy of license plate characters recognition based on K Nearest Neighbours (K-NN) classifier is presented in this paper. The system first prepares a contour form of the extracted character, then the angle and distance feature information about the character is extracted and finally K-NN classifier is used to character recognition. Angle and distance features of a character have been computed based on distribution of points on the bitmap image of character. In K-NN method, the Euclidean distance between testing point and reference points is calculated in order to find the k-nearest neighbours. We evaluated our method on the available dataset that contain 1200 sample. Using 70% samples for training, we tested our method on whole samples and obtained 99% correct recognition rate.Further, we achieved average 99.41% accuracy using three/strategy validation technique on 1200 dataset.

cs.CV↗
arXiv2014

Novel and Fast Algorithm for Extracting License Plate Location Based on Edge Analysis

Nowadays in developing or developed countries, the Intelligent Transportation System (ITS) technology has attracted so much attention to itself. License Plate Recognition (LPR) systems have many applications in ITSs, such as the payment of parking fee, controlling the traffic volume, traffic data collection, etc. This paper presents a new and fast method for license plate extraction based on edge analysis. our proposed method consist of four stage, which are edge detection, non-useable edge and noise removing, edge analysis and morphology-based license plate extraction. In the result part, the proposed algorithm is applied on vehicle database and the accuracy rate reached 98%. From the experimental results it is shown that the proposed method gives fairly acceptable level of accuracy for practical license plate recognition system.

cs.CV↗
arXiv2014

Classifiers fusion method to recognize handwritten persian numerals

Recognition of Persian handwritten characters has been considered as a significant field of research for the last few years under pattern analysing technique. In this paper, a new approach for robust handwritten Persian numerals recognition using strong feature set and a classifier fusion method is scrutinized to increase the recognition percentage. For implementing the classifier fusion technique, we have considered k nearest neighbour (KNN), linear classifier (LC) and support vector machine (SVM) classifiers. The innovation of this tactic is to attain better precision with few features using classifier fusion method. For evaluation of the proposed method we considered a Persian numerals database with 20,000 handwritten samples. Spending 15,000 samples for training stage, we verified our technique on other 5,000 samples, and the correct recognition ratio achieved approximately 99.90%. Additional, we got 99.97% exactness using four-fold cross validation procedure on 20,000 databases.

cs.CV↗
arXiv2014

Real-Time and Robust Method for Hand Gesture Recognition System Based on Cross-Correlation Coefficient

Hand gesture recognition possesses extensive applications in virtual reality, sign language recognition, and computer games. The direct interface of hand gestures provides us a new way for communicating with the virtual environment. In this paper a novel and real-time approach for hand gesture recognition system is presented. In the suggested method, first, the hand gesture is extracted from the main image by the image segmentation and morphological operation and then is sent to feature extraction stage. In feature extraction stage the Cross-correlation coefficient is applied on the gesture to recognize it. In the result part, the proposed approach is applied on American Sign Language (ASL) database and the accuracy rate obtained 98.34%.

cs.CV↗
arXiv2014

Real-Time Human-Computer Interaction Based on Face and Hand Gesture Recognition

At the present time, hand gestures recognition system could be used as a more expected and useable approach for human computer interaction. Automatic hand gesture recognition system provides us a new tactic for interactive with the virtual environment. In this paper, a face and hand gesture recognition system which is able to control computer media player is offered. Hand gesture and human face are the key element to interact with the smart system. We used the face recognition scheme for viewer verification and the hand gesture recognition in mechanism of computer media player, for instance, volume down/up, next music and etc. In the proposed technique, first, the hand gesture and face location is extracted from the main image by combination of skin and cascade detector and then is sent to recognition stage. In recognition stage, first, the threshold condition is inspected then the extracted face and gesture will be recognized. In the result stage, the proposed technique is applied on the video dataset and the high precision ratio acquired. Additional the recommended hand gesture recognition method is applied on static American Sign Language (ASL) database and the correctness rate achieved nearby 99.40%. also the planned method could be used in gesture based computer games and virtual reality.

cs.CV↗
arXiv2014

Real-Time and Efficient Method for Accuracy Enhancement of Edge Based License Plate Recognition System

License Plate Recognition plays an important role on the traffic monitoring and parking management. Administration and restriction of those transportation tools for their better service becomes very essential. In this paper, a fast and real time method has an appropriate application to find plates that the plat has tilt and the picture quality is poor. In the proposed method, at the beginning, the image is converted into binary mode with use of adaptive threshold. And with use of edge detection and morphology operation, plate number location has been specified and if the plat has tilt; its tilt is removed away. Then its characters are distinguished using image processing techniques. Finally, K Nearest Neighbour (KNN) classifier was used for character recognition. This method has been tested on available data set that has different images of the background, considering distance, and angel of view so that the correct extraction rate of plate reached at 98% and character recognition rate achieved at 99.12%. Further we tested our character recognition stage on Persian vehicle data set and we achieved 99% correct recognition rate.

cs.CV↗
arXiv2014

Novel and Tuneable Method for Skin Detection Based on Hybrid Color Space and Color Statistical Features

Skin detection is one of the most important and primary stages in some of image processing applications such as face detection and human tracking. So far, many approaches are proposed to done this case. Near all of these methods have tried to find best match intensity distribution with skin pixels based on popular color spaces such as RGB, CMYK or YCbCr. Results show these methods cannot provide an accurate approach for every kinds of skin. In this paper, an approach is proposed to solve this problem using statistical features technique. This approach is including two stages. In the first one, from pure skin statistical features were extracted and at the second stage, the skin pixels are detected using HSV and YCbCr color spaces. In the result part, the proposed approach is applied on FEI database and the accuracy rate reached 99.25 + 0.2. Further proposed method is applied on complex background database and accuracy rate obtained 95.40+0.31%. The proposed approach can be used for all kinds of skin using train stage which is the main advantages of it. Low noise sensitivity and low computational complexity are some of other advantages.

cs.CV↗
arXiv2014

New Method for Optimization of License Plate Recognition system with Use of Edge Detection and Connected Component

License Plate recognition plays an important role on the traffic monitoring and parking management systems. In this paper, a fast and real time method has been proposed which has an appropriate application to find tilt and poor quality plates. In the proposed method, at the beginning, the image is converted into binary mode using adaptive threshold. Then, by using some edge detection and morphology operations, plate number location has been specified. Finally, if the plat has tilt, its tilt is removed away. This method has been tested on another paper data set that has different images of the background, considering distance, and angel of view so that the correct extraction rate of plate reached at 98.66%.

cs.CV↗
arXiv2014

Novel and Automatic Parking Inventory System Based on Pattern Recognition and Directional Chain Code

The objective of this paper is to design an efficient vehicle license plate recognition System and to implement it for automatic parking inventory system. The system detects the vehicle first and then captures the image of the front view of the vehicle. Vehicle license plate is localized and characters are segmented. For finding the place of plate, a novel and real time method is expressed. A new and robust technique based on directional chain code is used for character recognition. The resulting vehicle number is then compared with the available database of all the vehicles so as to come up with information about the vehicle type and to charge entrance cost accordingly. The system is then allowed to open parking barrier for the vehicle and generate entrance cost receipt. The vehicle information (such as entrance time, date, and cost amount) is also stored in the database to maintain the record. The hardware and software integrated system is implemented and a working prototype model is developed. Under the available database, the average accuracy of locating vehicle license plate obtained 100%. Using 70% samples of character for training, we tested our scheme on whole samples and obtained 100% correct recognition rate. Further we tested our character recognition stage on Persian vehicle data set and we achieved 99% correct recognition.

cs.CV↗
arXiv2014

A robust and adaptable method for face detection based on Color Probabilistic Estimation Technique

Human face perception is currently an active research area in the computer vision community. Skin detection is one of the most important and primary stages for this purpose. So far, many approaches are proposed to done this case. Near all of these methods have tried to find best match intensity distribution with skin pixels based on popular color spaces such as RGB, HSI or YCBCR. Results show that these methods cannot provide an accurate approach for every kind of skin. In this paper, an approach is proposed to solve this problem using a color probabilistic estimation technique. This approach is including two stages. In the first one, the skin intensity distribution is estimated using some train photos of pure skin, and at the second stage, the skin pixels are detected using Gaussian model and optimal threshold tuning. Then from the skin region facial features have been extracted to get the face from the skin region. In the results section, the proposed approach is applied on FEI database and the accuracy rate reached 99.25%. The proposed approach can be used for all kinds of skin using train stage which is the main advantage among the other advantages, such as Low noise sensitivity and low computational complexity.

cs.CV↗
arXiv2014

Optimized Method for Iranian Road Signs Detection and recognition system

Road sign recognition is one of the core technologies in Intelligent Transport Systems. In the current study, a robust and real-time method is presented to identify and detect the roads speed signs in road image in different situations. In our proposed method, first, the connected components are created in the main image using the edge detection and mathematical morphology and the location of the road signs extracted by the geometric and color data; then the letters are segmented and recognized by Multiclass Support Vector Machine (SVMs) classifiers. Regarding that the geometric and color features ate properly used in detection the location of the road signs, so it is not sensitive to the distance and noise and has higher speed and efficiency. In the result part, the proposed approach is applied on Iranian road speed sign database and the detection and recognition accuracy rate achieved 98.66% and 100% respectively.

cs.CV↗