Shamik Tiwari

Bio: Shamik Tiwari is an academic researcher from University of Petroleum and Energy Studies. The author has contributed to research in topics: Computer science & Image restoration. The author has an hindex of 10, co-authored 48 publications receiving 339 citations. Previous affiliations of Shamik Tiwari include Mody University of Science & Technology.

Dermatologist-Level Classification of Skin Cancer Using Cascaded Ensembling of Convolutional Neural Network and Handcrafted Features Based Deep Neural Network

Abstract: Skin cancer is caused due to unusual development of skin cells and deadly type cancer. Early diagnosis is very significant and can avoid some categories of skin cancers, such as melanoma and focal cell carcinoma. The recognition and the classification of skin malignant growth in the beginning time is expensive and challenging. The deep learning architectures such as recurrent networks and convolutional neural networks (ConvNets) are developed in the past, which are proven appropriate for non-handcrafted extraction of complex features. To additional expand the efficiency of the ConvNet models, a cascaded ensembled network that uses an integration of ConvNet and handcrafted features based multi-layer perceptron is proposed in this work. This offered model utilizes the convolutional neural network model to mine non-handcrafted image features and colour moments and texture features as handcrafted features. It is demonstrated that accuracy of ensembled deep learning model is improved to 98.3% from 85.3% of convolutional neural network model.

Review of Motion Blur Estimation Techniques

Abstract: The goal of image restoration is to improve a given image in some predefined sense. Restoration attempts to recover an image by modelling the degradation function and applying the inverse process. Motion blur is a common type of degradation which is caused by the relative motion between an object and camera. Motion blur can be modeled by a point spread function consists of two parameters angle and length. Accurate estimation of these parameters is required in case of blind restoration of motion blurred images. This paper compares different approaches to estimate the parameters of a motion blur namely direction and length directly from the observed image with and without the influence of Gaussian noise. These estimated motion blur parameters can then be used in a standard nonblind deconvolution algorithm. Simulation results compare the performance of most common motion blur estimation methods.

Convolutional capsule network for COVID-19 detection using radiography images.

Abstract: Novel corona virus COVID-19 has spread rapidly all over the world. Due to increasing COVID-19 cases, there is a dearth of testing kits. Therefore, there is a severe need for an automatic recognition system as a solution to reduce the spreading of the COVID-19 virus. This work offers a decision support system based on the X-ray image to diagnose the presence of the COVID-19 virus. A deep learning-based computer-aided decision support system will be capable to differentiate between COVID-19 and pneumonia. Recently, convolutional neural network (CNN) is designed for the diagnosis of COVID-19 patients through chest radiography (or chest X-ray, CXR) images. However, due to the usage of CNN, there are some limitations with these decision support systems. These systems suffer with the problem of view-invariance and loss of information due to down-sampling. In this paper, the capsule network (CapsNet)-based system named visual geometry group capsule network (VGG-CapsNet) for the diagnosis of COVID-19 is proposed. Due to the usage of capsule network (CapsNet), the authors have succeeded in removing the drawbacks found in the CNN-based decision support system for the detection of COVID-19. Through simulation results, it is found that VGG-CapsNet has performed better than the CNN-CapsNet model for the diagnosis of COVID-19. The proposed VGG-CapsNet-based system has shown 97% accuracy for COVID-19 versus non-COVID-19 classification, and 92% accuracy for COVID-19 versus normal versus viral pneumonia classification. Proposed VGG-CapsNet-based system available at https://github.com/shamiktiwari/COVID19_Xray can be used to detect the existence of COVID-19 virus in the human body through chest radiographic images.

Statistical Moments based Noise Classification using Feed Forward Back Propagation Neural Network

Abstract: A neural network classification based noise identification method is presented by isolating some representative noise samples, and extracting their statistical features for noise type identification. The isolation of representative noise samples is achieved using prevalent used image filters whereas noise identification is performed using statistical moments features based classification system. The results of the experiments using this method show better identification of noise than those suggested in the recent works. General Terms Image denoising, Pattern recognition.

Flame Detection using Image Processing Techniques

Abstract: Dynamic textures are common in natural scenes Examples of dynamic textures in video include fire, smoke, trees in the wind, clouds, sky, ocean waves etc The fire is characterized using efficient features and detection of the same using a suitable processing Every pixel is checked for the presence or absence of fire using color features, and periodic behavior in fire regions is also analyzed In this paper we use combined approach of color detection, motion detection and area dispersion to detect fire in video data Firstly, the algorithm locates desired color regions in video frames, and then determines the region in the video where there is any movement, and in the last step we calculate the pixel area of the frame The combination of color, motion and area clues is used to detect fire in the video

WILDS: A Benchmark of in-the-Wild Distribution Shifts

Abstract: Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity, these real-world distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated collection of 8 benchmark datasets that reflect a diverse range of distribution shifts which naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training results in substantially lower out-of-distribution than in-distribution performance, and that this gap remains even with models trained by existing methods for handling distribution shifts. This underscores the need for new training methods that produce models which are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at this https URL.

Extracting Training Data from Large Language Models

Abstract: It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. We demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. We comprehensively evaluate our extraction attack to understand the factors that contribute to its success. Worryingly, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models.

Using Deep Learning for Classification of Lung Nodules on Computed Tomography Images.

Abstract: Lung cancer is the most common cancer that cannot be ignored and cause death with late health care. Currently, CT can be used to help doctors detect the lung cancer in the early stages. In many cases, the diagnosis of identifying the lung cancer depends on the experience of doctors, which may ignore some patients and cause some problems. Deep learning has been proved as a popular and powerful method in many medical imaging diagnosis areas. In this paper, three types of deep neural networks (e.g., CNN, DNN, and SAE) are designed for lung cancer calcification. Those networks are applied to the CT image classification task with some modification for the benign and malignant lung nodules. Those networks were evaluated on the LIDC-IDRI database. The experimental results show that the CNN network archived the best performance with an accuracy of 84.15%, sensitivity of 83.96%, and specificity of 84.32%, which has the best result among the three networks.

Contrastive Learning of Medical Visual Representations from Paired Images and Text

Abstract: Learning visual representations of medical images is core to medical image understanding but its progress has been held back by the small size of hand-labeled datasets. Existing work commonly relies on transferring weights from ImageNet pretraining, which is suboptimal due to drastically different image characteristics, or rule-based label extraction from the textual report data paired with medical images, which is inaccurate and hard to generalize. We propose an alternative unsupervised strategy to learn medical visual representations directly from the naturally occurring pairing of images and textual data. Our method of pretraining medical image encoders with the paired text data via a bidirectional contrastive objective between the two modalities is domain-agnostic, and requires no additional expert input. We test our method by transferring our pretrained weights to 4 medical image classification tasks and 2 zero-shot retrieval tasks, and show that our method leads to image representations that considerably outperform strong baselines in most settings. Notably, in all 4 classification tasks, our method requires only 10% as much labeled training data as an ImageNet initialized counterpart to achieve better or comparable performance, demonstrating superior data efficiency.

Detection of brain tumor in MRI images, using combination of fuzzy c-means and SVM

Abstract: MRI is the most important technique, in detecting the brain tumor. In this paper data mining methods are used for classification of MRI images. A new hybrid technique based on the support vector machine (SVM) and fuzzy c-means for brain tumor classification is proposed. The purposed algorithm is a combination of support vector machine (SVM) and fuzzy c-means, a hybrid technique for prediction of brain tumor. In this algorithm the image is enhanced using enhancement techniques such as contrast improvement, and mid-range stretch. Double thresholding and morphological operations are used for skull striping. Fuzzy c-means (FCM) clustering is used for the segmentation of the image to detect the suspicious region in brain MRI image. Grey level run length matrix (GLRLM) is used for extraction of feature from the brain image, after which SVM technique is applied to classify the brain MRI images, which provide accurate and more effective result for classification of brain MRI images.