ATAnik Tahabilder

Anik Tahabilder

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Anik Tahabilder has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A deep learning approach for COVID-19 and pneumonia detection from chest X-ray images

    by Ahmmad Musha, Abdullah Al Mamun, Anik Tahabilder, Jakir Hossen, Busrat Hossen, Sima Ranjbari

    There has been a surge in biomedical imaging technologies with the recent advancement of deep learning. It is being used for diagnosis from X-ray, computed tomography (CT) scan, electrocardiogram (ECG), and electroencephalography (EEG) images. However, most of them are solely for particular disease detection. In this research, a computer-aided deep learning model named COVID-CXDNetV2 has been presented to detect two separate diseases, coronavirus disease 2019 (COVID-19) and pneumonia, from the X-ray images in real-time. The proposed model is made based on you only look once (YOLOv2) with residual neural network (ResNet) and trained by a vast X-ray images dataset containing 3788 samples of three classes named COVID-19 pneumonia and normal. The model has obtained the maximum overall classification accuracy of 97.9% with a loss of 0.052 for multiclass classification (COVID-19, pneumonia, and normal) and 99.8% accuracy, 99.52% sensitivity, 100% specificity with a loss of 0.001 for binary classification (COVID-19 and normal), which beats some current state-of-the-art results. Authors believe that this method will be applicable in the medical domain for the diagnosis and will significantly contribute to real life.


  • Open Access

    Acute lymphoblastic leukemia detection approach from peripheral blood smear using color threshold and morphological techniques

    by Abdullah Al Mamun, Jakir Hossen, Anik Tahabilder, Ahmmad Musha, Rehnuma Hasnat, Sohag Kumar Saha

    Acute lymphoblastic leukemia (ALL) has recently been one of the most significant concerns in cancers, especially child and old age. Therefore, crying needs to diagnose leukemia as early as possible, increasing the treatment options and patient survivability. Some basic handicraft leukemia detection processes have been introduced in this arena though these are not so accurate and efficient. The proposed approach has been introduced an automated ALL recognition system from the peripheral blood smear. Initially, the color threshold has been applied to segment lymphocytes blood cells from the blood smear. Some post-processing techniques like morphological operation and watershed have been executed to segment the particular lymphocytes cell. Finally, we used a support vector machine (SVM) classifier to classify the cancerous image frames using a statistical feature vector obtained from the segmented image. The proposed framework has achieved the highest accuracy of 99.21%, the sensitivity of 98.45%, specificity of 99%, the precision of 99%, and F1 score of 99.1%, which has beat existing and common states of art methods. We are confident that the proposed approach will positively impact the ALL detection arena.


  • Open Access

    Efficient lane marking detection using deep learning technique with differential and cross-entropy loss

    by Abdullah Al Mamun, Poh Ping Em, Jakir Hossen, Anik Tahabilder, Busrat Jahan

    Nowadays, researchers are incorporating many modern and significant features on advanced driver assistance systems (ADAS). Lane marking detection is one of them, which allows the vehicle to maintain the perspective road lane. Conventionally, it is detected through handcrafted and very specialized features and goes through substantial post-processing, which leads to high computation, and less accuracy. Additionally, this conventional method is vulnerable to environmental conditions, making it an unreliable model. Consequently, this research work presents a deep learning-based model that is suitable for diverse environmental conditions, including multiple lanes, different daytime, different traffic conditions, good and medium weather conditions, and so forth. This approach has been derived from plain encode-decode E-Net architecture and has been trained by using the differential and cross-entropy losses for the backpropagation. The model has been trained and tested using 3,600 training and 2,700 testing images from TuSimple, a robust public dataset. Input images from very diverse environmental conditions have ensured better generalization of the model. This framework has reached a max accuracy of 96.61%, with an F1 score of 96.34%, a precision value of 98.91%, and a recall of 93.89%. Besides, this model has shown very small false positive and false negative values of 3.125% and 1.259%, which bits the performance of most of the existing state of art models.


  • Open Access

    A Deep Learning Approach for Lane Markings Detection Applying Encode-Decode Instant Segmentation Network

    by Abdullah Al Mamun, Em Poh Ping, Jakir Hossen, Anik Tahabilder

    Numerous people have to suffer from disability and death due to unintentional road accidents, resulting in the loss of a significant amount of financial assets. As a result, several essential features of Advanced Driver Assistance Systems are incorporated by researchers. Lane marking detection (LMD) is a fundamental ADAS technology that helps the vehicle to keep its position in the road lane. The current study, which employs Deep Learning (DL) methodologies, has several research constraints due to various problems. Researchers sometimes encounter difficulties in LMD due to environmental factors such as the variant of lights, obstacles, shadow, and curve lanes. As a result, this study presented the Encode-Decode Instant Segmentation Network (EDIS-Net) as a DL methodology for detecting lane markings under various environmental situations with high-performance outcomes. The framework is based on the E-Net architecture incorporating a combination of discriminative and cross-entropy losses. The encoded segment was split into binary and instant segmentation to extract information about the lane pixels and pixel position. Densely-Based Spatial Clustering of Application with Noise is employed to connect the predicted lane pixels and get the final output. The system was trained with data augmentation on the Tusimple dataset and then tested on three datasets: Tusimple, CalTech, and a local dataset. On the Tusimple dataset, the model achieved 97.39 % accuracy. Furthermore, it has an average accuracy of 97.07 % and 96.23 % on the CalTech and local datasets. On testing datasets, the EDIS-Net exhibited promising results when compared to existing LMD approaches. Because the proposed framework performs better on the testing datasets, it can be argued that it can recognise lane markings well in various scenarios. This study presented a novel EDIS-Net technique for efficiently detecting lane markings. It also includes the performance verification tests in two public datasets, including a local dataset.

  • Support Directional Shifting Vector: A Direction Based Machine Learning Classifier

    by Kowsher, Imran Hossen, Anik Tahabilder, Nusrat Jahan Prottasha, Kaiser Habib, Zafril Rizal M. Azmi

    Support Directional Shifting Vector: A Direction Based Machine Learning Classifier


  • Open Access

    BnVec: Towards the Development of Word Embedding for Bangla Language Processing

    by Kowsher, Jashim Uddin, Anik Tahabilder, Nusrat Jahan Prottasha, Mahid Ahmed, K. M. Rashedul Alam, Tamanna Sultana

    Progression in machine learning and statistical inference are facilitating the advancement of domains like computer vision, natural language processing (NLP), automation & robotics, and so on. Among the different persuasive improvements in NLP, word embedding is one of the most used and revolutionary techniques. In this paper, we manifest an open-source library for Bangla word extraction systems named BnVec which expects to furnish the Bangla NLP research community by the utilization of some incredible word embedding techniques. The BnVec is splitted up into two parts, the first one is the Bangla suitable defined class to embed words with access to the six most popular word embedding schemes (CountVectorizer, TF-IDF, Hash Vectorizer, Word2vec, fastText, and Glove). The other one is based on the pre-trained distributed word embedding system of Word2vec, fastText, and GloVe. The pre-trained models have been built by collecting content from the newspaper, social media, and Bangla wiki articles. The total number of tokens used to build the models exceeds 395,289,960. The paper additionally depicts the performance of these models by various hyper-parameter tuning and then analyzes the results.