Leukemia is a fatal disease of white blood cells which affects the blood and bone marrow in human body. We deployed deep convolutional neural network for automated detection of acute lymphoblastic leukemia and classification of its subtypes into 4 classes, that is, L1, L2, L3, and Normal which were mostly neglected in previous literature. In contrary to the training from scratch, we deployed pretrained AlexNet which was fine-tuned on our data set. Last layers of the pretrained network were replaced with new layers which can classify the input images into 4 classes. To reduce overtraining, data augmentation technique was used. We also compared the data sets with different color models to check the performance over different color images. For acute lymphoblastic leukemia detection, we achieved a sensitivity of 100%, specificity of 98.11%, and accuracy of 99.50%; and for acute lymphoblastic leukemia subtype classification the sensitivity was 96.74%, specificity was 99.03%, and accuracy was 96.06%. Unlike the standard methods, our proposed method was able to achieve high accuracy without any need of microscopic image segmentation.

Leukaemia is a form of blood cancer which affects the white blood cells and damages the bone marrow. Usually complete blood count (CBC) and bone marrow aspiration are used to diagnose the acute lymphoblastic leukaemia. It can be a fatal disease if not diagnosed at the earlier stage. In practice, manual microscopic evaluation of stained sample slide is used for diagnosis of leukaemia. But manual diagnostic methods are time-consuming, less accurate, and prone to errors due to various human factors like stress, fatigue, and so forth. Therefore, different automated systems have been proposed to wrestle the glitches in the manual diagnostic methods. In recent past, some computer-aided leukaemia diagnosis methods are presented. These automated systems are fast, reliable, and accurate as compared to manual diagnosis methods. This paper presents review of computer-aided diagnosis systems regarding their methodologies that include enhancement, segmentation, feature extraction, classification, and accuracy.

Autonomous path following has gained tremendous popularity during the last few decades. Numerous researchers have contributed to the development of highly automated navigation systems using different types of sensors and their combination. However, their proposed approaches do not provide a cost-efficient solution because of the deployment of exorbitant and sophisticated sensors, which remains a challenging problem for customized vehicles used in academic research. To overcome this issue, this study presents an economically efficient sensorless steering angle approach that employs a single camera for steering control and quick response (QR) based localization of a vehicle. Moreover, we used SONAR for object detection in a defined route to avoid possible collisions. The proposed technique combines a probabilistic Hough transform for lane detection and QR codes, which helps the vehicle stay in its lane for stabilized control. To prove the efficiency of our approach, we tested it on our developed prototype vehicle named EMO. To validate the proposed approach through in-field testing, we designed a customized test track within the campus. The experimental results show the benefit of our proposed approach compared to existing methods available in the literature.

In an automated steering system of the self-driving vehicles, the steering wheel angle is measured by the absolute angular displacement sensors or relative angle sensors. However, these sensors either encompass global navigation satellite systems (GNSS)/gyroscope – Micro Electromechanical-Sensor (MEMS) based solutions or comprise of the complex gear-based mechanical structure which results in latency and additive bias in the accumulative steering angle assessment. To address these issues, we propose a novel steering angle assessment system based on enhanced gear mechanism along with the adapted rotation paradigm for the customized self-driving vehicles. Additionally, a digital signal processing system has been introduced to resolve the issues in the identification of absolute central and max-bounding steering wheels position in self-driving vehicles. In assistance with the proposed mechanism, an algorithm has also been proposed to optimize the computed steering angle to minimalize the effect of additive bias in the accuracy. The proposed mechanism has been installed in the customized self-driving testbed vehicle and rigor validation has been performed in the straight and curvy road scenarios. Finally, the comparison study has been carried out between the conventional relative sensor and the proposed mechanism to show the accuracy and effectiveness of the proposed mechanism in terms of error rate, stability, and deviation.

In the last couple of decades, autonomous human assistance robots have been enormously attracting the industrial sector. For this purpose, numerous researchers have contributed towards designing efficient and robust human assistance mechanisms. However, their proposed approaches do not provide a cost-effective solution due to the deployment of exorbitant sensors and sophisticated infrastructure. Besides, it was quite challenging for existing human-following robots to track their assigned human companion in different illusional states and luminous conditions while detecting obstacles and taking respective maneuvers (i.e., abrupt turns, etc.). Moreover, self-driving solutions need to take fast and real-time actions to avoid collisions in the designated environment. For this purpose, literature has shown the efficiency of YOLOv3 with respect to providing real-time results in latency-sensitive applications. Hence, to overcome this dilemma, we propose to develop an economically efficient deep learning- based smart labor assistance trolley that uses the YOLO v3 as a core detective deep learning technique with advance and efficient perception and motion planning modules. The perception module succours the autonomous trolley to precisely detect and classify the objects. While the motion planning module uses the specific intended target detection technique to follow the targeted person in crowded environment. These techniques make the autonomous trolley able to take expeditious, meticulous, and conspicuous action in real-time. The labor assistant robot detects and tracks the respective person using YOLOv3. To validate the efficiency of the proposed solution, we have performed a series of experiments considering different test cases. Our proposed work achieved a mean average precision of 0.81%.

The advancement of self-driving vehicles' ability to perceive their environment is highly dependent on computer vision-based object detection and recognition. Over the past few decades, many substantial object detection and recognition datasets have been proposed, including the Karlsruhe Institute of Technology and Toyota Technological (KITTI) dataset, the Beijing Institute of Technology (BIT) vehicle dataset, etc. However, these datasets consist of smooth and organised urban road traffic of Western nations, ignoring the congested and disordered traffic of Asian nations (i.e., Pakistan, India, etc.). To overcome the above-mentioned issues, a substantial dataset named Control Automotive and Robotics Lab- Object Detection Dataset (CARL-ODD) has been proposed for self-driving vehicles to navigate safely while having a detailed perception of their surroundings in an Asian environment. CARL-ODD dataset was collected from over eighty cities of Pakistan containing highways, motorways, and congested as well as diverse traffic scenarios of urban, rural, and hilly areas. The use of both mono and stereo-vision-based driving videos in the dataset provided a considerable advantage over the state-of-the-art datasets. The collected data were then labelled, and 20,000 representative images were selected for the perception module. Moreover, to establish a baseline for CARL-ODD, we have also compared the various pre-trained single-stage and multistage object detection and recognition algorithms with the dataset. These pre-trained object detection algorithms, such as YOLO and Faster R-CNN, have been implemented to detect and classify various on-road vehicles present in the Asian environment. This approach can save time and resources compared to training a model from scratch. The major findings as a result of the analysis showed that the CARL-ODD dataset demonstrated strong performance in detecting and recognising surrounding vehicles, while achieving 91% average precision for single-stage detectors and 93% average precision for multistage object detectors. This work outperforms the state-of-the-art datasets in terms of vehicle classes and diverse traffic scenarios.