Artificial Intelligence Projects Ideas & Topics

Image blur detection and removal have been one of the major topics of research in image processing in the recent years. These blur detection and removal algorithms have many real world applications including image restoration and image enhancement. Image blur can include motion blur and out-of-focus blur or blurring due to lens imperfections. This paper also proposes an approach for motion blur detection and removal involving Convolutional Neural Network(CNN) and Generative Adversarial Network(GAN).

Having diseases is quite natural in crops due to changing climatic and environmental conditions. Diseases affect the growth and produce of the crops and often difficult to control. To ensure good quality and high production, it is necessary to have accurate disease diagnosis and control actions to prevent them in time. Grape which is widely grown crop in India and it may be affected by different types of diseases on leaf, stem and fruit. Fruit diseases which are the early symptoms caused due to Gray_mold, Healthy Grapes, powdery_mildew and Sour rot. So, there is a need to have an automatic system that can be used to detect the type of diseases and to take appropriate actions suggest remedies.

Now days difficult find drone are bird because drones are popular, and increasingly everywhere, so we implemented this project detect the drones. In addition to their useful applications, an alarming concern regarding physical infrastructure security, safety, and surveillance at airports has arisen due to the potential of their use in malicious activities. In recent, there have been many reports of the unauthorized use of various types of drones at airports and the disruption of airline operations and some human cannot find the drone or birds. This problem, that’s why we implement this project using deep learning-based method for the efficient detection and recognition of two types of drones and birds. Evaluation of the proposed approach with the prepared image dataset demonstrates better efficiency compared to existing detection systems.. Furthermore, drones are often confused with birds because of their physical and behavioral similarity. The proposed method is not only able to detect the presence or absence of drones in an area but also to recognize and distinguish between two types of drones, as well as distinguish them from birds. The dataset used in this work to train the network consists of some visible images containing two types of drones and birds. The proposed deep learning method can directly detect and recognize two types of drones and birds.

We using Deep learning algorithms and has been applied for image detection and classification, with good results in the medicine such as medical image analysis. This paper aims to support the detection of barin hemorrhage in computed tomography (CT) images using deep learning algorithms and convolutional neural networks (CNN). The motivation of this work is the difficulty of physicians when they face the task to identify brain hemorrhage, but this project easily find out disease especially when they are in the primary stages of brain bleeding, making a misdiagnosis. A some of CT studies were used to train and evaluate two convolutional neuronal networks in the task of classifying hemorrhage or non-hemorrhage. The proposed CNN networks reach 92% of accuracy.

Today’s modern vehicles contain anywhere from sixty to one-hundred sensors and exhibit the characteristics of Cyber-Physical- Systems (CPS). There is a high degree of coupling, cohesiveness, and interactions among vehicle’s CPS components (e.g., sensors, devices, systems, systems-of-systems) across sensing, communication, and control layers. Cyber-attacks in the sensing or communication layers can compromise the security of the control layer. This paper provides a detailed review of potential cyber threats related to the sensing layer. Notably, the focus is mainly towards two categories of sensors: vehicle dynamics sensors (e.g., Tire Pressure Monitoring Systems (TPMS), magnetic encoders, and inertial sensors) and environment sensors (e.g., Light Detection and Ranging (LiDAR), ultrasonic, camera, Radio Detection and Ranging (Radar) systems, and Global Positioning System (GPS) units). The paper also offers perspectives through existing countermeasures from literature and stresses the need for data-driven cybersecurity solutions.

The number of Wireless sensor network (WSN) deployments have been growing so exponentially over the recent years.Due to their small size and cost-e_ective, WSN are attracting many industries to use them in various applications.

Environmental monitoring, security of buildings and precision agriculture are few example among several other _elds.However, WSN faces high security threats considering most of them are deployed in unattended nature and hostile environment.

In the aim of providing secure data processing in the WSN, many techniques are proposed to protect the data privacy while being transferred from the sensors to the base station.This work is focusing on attack detection which is an essential task to secure the network and the data.

Anomaly detection is a key challenge in order to ensure the security and prevent malicious attacks in wireless sensor networks.Various machine learning techniques have been used by researchers these days to detect anomalies using o_inelearning algorithms.

On the other hand online learning classi_ers have not been thoroughly addressed in the literature. Our aim is to provide an intrusion detection model compatible with the characteristics of WSN. This model is built based on information gain ratio and the online Passive aggressive classi_er.

Firstly, the information gain ratio is used to select the relevant features of the sensor data. Secondly, the online Passive aggressive algorithm is trained to detect and classify di_erent type of Deny of Service attacks. The experiment was conducted on a wireless sensor network-detection system (WSN-DS) dataset.

The proposed model ID-GOPA results detection rate of 96% determining whether the network is in its normal mode or exposed to any type of attack.The detection accuracy is 86%, 68%, 63%, and 46% for scheduling, grayhole, ooding and blackhole attacks, respectively, in addition to 99% for normal tra_c.

These results shows that our model based on o_ine learning can be providing good anomaly detection to the WSN and replace online learning in some cases.

Medicinal Plants and Natural Medicine are becoming more important in our lives. It will be very useful to know so many remedies for better health. Many people already use these techniques and enjoy day to day living a fuller life. There are a number of advantages associated with the use of medicinal plants rather than pharmaceuticals such as side effects or lower cost. Hence, this model proposes an AI-based solution designed to provide in-depth information on medicinal leaves and herbs with just a single image. This model will help people living in remote areas to understand the medicinal benefits of trees and plants around them and harness them for their well-being. This work instantly determines the identification of plants with Object detection technology in AI and gives information of the medicinal leaf about medicinal virtues on health. It shows the virtues, scientific names, chemical compounds and how to use plants to have a natural remedy. It also gives the detailed information about the medicinal leaf if it has any special precautions to be taken in the cases of pregnant and breastfeeding women to avoid problems such as miscarriages etc . There are several plants that look alike from afar and it's typically little details that distinguish 2 species of a similar genus. Flowers, fruits and leaves are the foremost characteristic organs of a species and it is them that ought to be photographed first. However the other details are often helpful to identify plants are such as thorns, buds or hair on the stem.

Agriculture is India's economic backbone. It is a source of income for roughly half of India's population. It is mostly a produce-oriented business; hence the profit or loss is determined by the yield gained. The most pressing concerns in agriculture is controlling weeds developing amid plantation crops. Weeds are found on every farmland in the globe [1].

Weeds are plants that grow unintentionally in agricultural soils and are undesirable. These compete for light, water, and nutrients with attractive plants, introduce illnesses or viruses, and attract harmful insects and pests, resulting in yield loss[2].

Weeds don't have a botanical categorization since a plant that is a weed in one place isn't a weed in another. It's used to describe any plant that grows or reproduces aggressively, or that is growing outside of its natural environment[3].

Outside of the plant world, the phrase is sometimes used to generically characterize organisms that can exist in a variety of settings and reproduce swiftly. These have seeds that last for many years in the soil seed bank. In agricultural research, weed categorization is a severe problem. Identification of weed species for management requires weed categorization. Weeds are divided into the grass and broadleaf groups depending on the shape of their leaves. Because grass and broadleaf weeds are treated differently owing to the selectivity of particular herbicides for the specific group, the division between these two classifications is appropriate[4].

In intracranial hemorrhage treatment patient mortality depends on prompt diagnosis based on a radiologist’s assessment of CT scans. In this paper, we investigate the intracranial hemorrhage detection problem and built a deep learning model to accelerate the time used to identify the hemorrhages. To assist with this process, a deep learning model can be used to accelerate the time it takes to identify them. In particular, we built a convolutional neural network based on ResNet for the classification. Using 752,803 DICOM files collected from four international universities by the Radiological Society of North America (RSNA) [1], we trained and tested a ResNet- 50 based model for predicting the hemorrhage type. Our model has an accuracy of 93.3% in making the correct multiclass prediction and an average per-class recall score of 76%. We show it is possible to achieve an average recall of 86% while maintaining 70% precision via tuning the prediction thresholds.

With the increasing number of drones, the danger of their illegal use has become elevant. This has necessitated the creation of automatic drone protection systems. One of the important tasks solved by these systems is the reliable detection of drones near guarded objects. This problem can be solved using various methods. From the point of view of the price–quality ratio, the use of video cameras for a drone detection is of great interest. However, drone detection using visual information is hampered by the large similarity of drones to other objects, such as birds or airplanes.

Face alignment is a crucial component in most face analysis systems. It focuses on identifying the location of several keypoints of the human faces in images. Although several methods and models are available to developers in popular computer vision libraries, they still struggle with challenges such as insufficient illumination, extreme head poses, or occlusions, especially when they are constrained by the needs of real-time applications. Throughout this article, we propose a set of training strategies and implementations based on data augmentation, software optimization techniques that help in improving a large variety of models belonging to several real-time algorithms for face alignment. We propose a method to detect the facial landmarks. We are using using CNN algorithm. We apply CNN algorithm for the prediction of the face landmarks. Hence Machine Learning is used to implement the proposed system.

The rampant coronavirus disease 2019 (COVID-19) has brought global crisis with its deadly spread to more than 180 countries, and about 3,519,901 confirmed cases along with 247,630 deaths globally as on May 4, 2020. The absence of any active therapeutic agents and the lack of immunity against COVID19 increases the vulnerability of the population.

Since there are no vaccines available, social distancing is the only feasible approach to fight against this pandemic. Motivated by this notion, this article proposes a deep learning based framework for automating the task of monitoring social distancing using surveillance video.

The proposed framework utilizes the YOLO object detection model to segregate humans from the background and to track the identified people with the help of bounding boxes The violation index term is proposed to quantize the non adoption of social distancing protocol. From the experimental analysis, it is observed that the YOLO with Deepsort tracking scheme displayed best results with balanced criteria to monitor the social distancing in real-time.

Parkinson's disease is a progressive and chronic neurodegenerative disorder. As the dopamine-generating neurons in parts of the brain become damaged or die, people begin to experience difficulty in speaking, writing, walking, or completing other simple tasks. These symptoms grow worse over time, thus resulting in the increase of its severity in patients. In this paper, we have proposed a methodology for the prediction of Parkinson’s disease severity using deep neural networks on UCI’s Parkinson’s Telemonitoring Voice Data Set of patients. We have used ‘TensorFlow’ deep learning library of python to implement our neural network for predicting the severity. The accuracy values obtained by our method are better as compared to the accuracy obtained in previous research work.

This paper presents facial detection and emotion analysis software developed by and for secondary students and teachers. The goal is to provide a tool that reduces the time teachers spend taking attendance while also collecting data that improves teaching practices. Disturbing current trends regarding school shootings motivated the inclusion of emotion recognition so that teachers are able to better monitor students’ emotional states over time. This will be accomplished by providing teachers with early warning notifications when a student significantly deviates in a negative way from their characteristic emotional profile. This project was designed to save teachers time, help teachers better address student mental health needs,and motivate students and teachers to learn more computer science, computer vision, and machine learning as they use and modify the code in their own classrooms. Important takeaways from initial test results are that increasing training images increases the accuracy of the recognition software, and the farther away a face is from the camera, the higher the chances are that the face will be incorrectly recognized.

Face Detection has evolved as a very popular problem in Image processing and Computer Vision. Many new algorithms are being devised using convolutional architectures to make the algorithm as accurate as possible. These convolutional architectures have made it possible to extract even the pixel details. We aim to design a binary face classifier which can detect any face present in the frame irrespective of its alignment. We present a method to generate accurate face segmentation masks from any arbitrary size input image. Beginning from the RGBimage of any size, the method uses Predefined Training Weights of VGG – 16 Architecture for feature extraction. Training is performed through Fully Convolutional Networks to semantically segment out the faces present in that image. Gradient Descent is used for training while Binomial Cross Entropy is used as a loss function. Further the output image from the FCN is processed to remove the unwanted noise and avoid the false predictions if any and make bounding box around the faces. Furthermore, proposed model has also shown great results in recognizing non-frontal faces. Along with this it is also able to detect multiple facial masks in a single frame. Experiments were performed on Multi Parsing Human Dataset obtaining mean pixel level accuracy of 93.884 % for the segmented face masks.

Age and gender identification have become a major part of the network, security and care. It has a common use in age specific content access for children. Social media uses it in delivering layered ads and marketing to extend it’s a reach. Face recognition has developed to a great extent that we have to map it further in getting more useful results having different approaches. In this paper, we propose deep CNN to improve age and gender predication from significant results can be obtained and a significant improvement can be seen in various tasks such as face recognition. A simple convolutional network architecture is proposed to make a noticeable improvement in this field using existing methods. Using deep CNN, model is trained to an extent that accuracy of Age and Gender become 79% using HAAR Feature-based Cascade Classifiers is an effective method proposed by Paul Viola and Michael Jones. It is a machine learning based approach where a cascade function is trained from a lot of positive and negative images. It is then used to detect objects in other images.

In Artificial Intelligence (AI), the contents of an image are generated automatically which involves computer vision and NLP (Natural Language Processing). The neural model which is regenerative, is created. It depends on computer vision and machine translation. This model is used to generate natural sentences which eventually describes the image. This model consists of Convolutional Neural Network(CNN) as well as Recurrent Neural Network(RNN). The CNN is used for feature extraction from image and RNN is used for sentence generation. The model is trained in such a way that if input image is given to model it generates captions which nearly describes the image. The accuracy of model and smoothness or command of language model learns from image descriptions is tested on different datasets. These experiments show that model is frequently giving accurate descriptions for an input image.

The COVID-19 is an unparalleled crisis leading to huge number of casualties and security problems. In order to reduce the spread of coronavirus, people often wear masks to protect themselves. This makes the face recognition a very dicult task since certain parts of the face are hidden. A primary focus of researchers during the ongoing coronavirus pandemic is to come up with suggestions to handle this problem through rapid and ecient solutions. In this paper, we propose a reliable method based on discard masked region and deep learning based features in order to address the problem of masked face recognition process. The rst step is to discard the masked face region. Next, we apply a pre-trained deep Convolutional neural networks (CNN) to extract the best features from the obtained regions (mostly eyes and forehead regions). Finally, the Bag-of-features paradigm is applied on the feature maps of the last convolutional layer in order to quantize them and to get a slight representation comparing to the fully connected layer of classical CNN. Finally, MLP is applied for the classication process. Experimental results on Real-World-Masked-Face-Dataset show high recognition performance

Parkinson's disease (PD) is a neuro degenerative disorder, which is responsible for the deterioration of motor function due to loss of dopamine-producing brain cells i.e.neurons. Tremors, stiffness, slowness in movements, shaking, and impaired balance are some of the primary symptoms of PD. In this paper, two neural network based models namely, VGFR Spectrogram Detector and Voice Impairment Classifier have been introduced, which aim to help doctors and people in diagnosing disease at an early stage. An extensive empirical evaluation of CNNs (Convolutional Neural Networks) has been implemented on large-scale image classification of gait signals converted to spectrogram images and deep dense ANNs (Artificial Neural Networks) on the voice recordings, to predict the disease. The experimental results indicate that the proposed models of artificial neural networks outperformed the existing state of the arts in terms of accuracy.The classification accuracy on VGFR Spectrogram Detector is recorded as 88.1% while Voice Impairment Classifier has shown 89.15% accuracy.

Image-based fruit calorie estimation is crucial to diverse mobile applications for recording everyday meal. However, some of them need human help for calorie estimation, and even if it is automatic, fruit categories are often limited or images from multiple viewpoints are required. Then, it is not yet achieved to estimate fruit calorie with practical accuracy and estimating fruit calories from a fruit photo is an unsolved problem. Therefore, in this paper, we propose estimating fruit calorie from a fruit photo by simultaneous learning of fruit calories using deep learning. We present a system which can recognize the fruit image, and then predict its nutritional contents, such as calories. We present CNN-based approaches to these problems, with promising preliminary results.

Alzheimer’s Disease is a progressive and irreversible neurological disease and is the most common cause of Dementia in people of the age 65 years and above. Detection of Alzheimer’s disease at prodromal stage is very important as it can prevent serious damage to the patient’s brain. In this paper, a method to detect Alzheimer’s Disease from MRI using Machine Learning approach. The MRI image dataset is used in the algorithm. The proposed approach detects the various stages of Alzheimer’s Disease such as moderate-demented and non-demented using CNN algorithm.

Pneumonia is a dangerous overwhelming illness influencing one or the two lungs in people reliably accomplished by minute living things called Streptococcus pneumoniae. One of every three passings in India is caused because of pneumonia as detailed by World Health Organization (WHO). Chest X-Rays which are utilized to research pneumonia need pro radiotherapists for examination. In this manner, building up a modified structure for perceiving pneumonia would be useful for treating the malady quickly especially in remote zones. Considering the accomplishment of noteworthy learning calculations in isolating clinical pictures, Convolutional Neural Networks (CNNs) have extended a lot of thought for sullying plan. Likewise, highlights recognized by pre-masterminded CNN models for epic degree datasets are a lot of significant in picture blueprint assignments. In this work, we evaluate the handiness of pre-masterminded CNN models used as highlight extractors followed by various classifiers for the solicitation for remarkable and standard chest X-Rays. We intentionally pick the ideal CNN model for the clarification. Quantifiable outcomes picked up shows that pretrained CNN models utilized close by oversaw classifier estimations can be critical in isolating chest X-shaft pictures, explicitly to see Pneumonia.

Handwritten digit recognition has recently been of very interest among the researchers because of the evolution of various Machine Learning, Deep Learning and Computer Vision algorithms. In this report, I compare the results of some of the most widely used Machine Learning Algorithms like SVM, KNN & RFC andwith Deep Learning algorithm like multilayer CNN using Keras with Theano and Tensorflow. Using these, I was able to get the accuracy of 98.70% using CNN (Keras+Theano) as compared to 97.91% using SVM, 96.67% using KNN, 96.89% using RFC

The automatic analysis and understanding of music by the computer is the new possibility in the field of music information retrieval [5]. Since the diversity and richness of music content, are high, a multitude of research topics in this field are pursued by the researchers, ranging from computer science, digital signal processing, mathematics, and statistics applied to musicology. Recent development in music information retrieval include automatic audio genre/mood classification, music similarity computation, audio artist identification, audio-to- score alignment, query-by-singing/humming and so on [7]. Content-based music recommendation is one the feasible application that can be provided. From the context information, we can achieve more intelligent context-based music recommendation. Multidisciplinary efforts such as emotion description, emotion detection/recognition [4], feature based classification, and inference-based recommendation are needed for the achievement in content based music recommendation system.

Over the last few years, Chatbots have played a prominent role as human-computer interfaces. Chatbots are generally composed of three modules: the user interface, an interpreter, and a knowledge base. Chatbot can be defines as a program that attempts to simulate typed conversation, with the aim of at least temporarily fooling the human into thinking they were talking to another person. Basically, chatbot is a conversational agent that can interact with user in a given subject using the natural language. Many chatbots have been deployed on the internet for the purpose of education, customer service site, guidance, entertainment. A bot is a software application that performs automated task and chatbots come under the category of bots that live in various chat platforms. A chatbot can converse with humans so the idea of conversation is primary to a chatbot. In other words, a chatbot is a computer program that can converse with humans using artificial intelligence in messaging platforms.

The latest generation of convolutional neural networks (CNNs) has achieved impressive results in the field of image classification. This paper is concerned with a new approach to the development of plant disease recognition model, based on leaf image classification, by the use of deep convolutional networks. Novel way of training and the methodology used facilitate a quick and easy system implementation in practice. The developed model is able to recognize three types of diseases of one plant and pesticides and/or fertilizers are advised according to the severity of the diseases. The type of green leaf diseases is recognized by CNN and KNN algorithms. After recognition, the predictive remedy is suggested that can help agriculture related people and organizations to take appropriate actions against these diseases.

The latest generation of convolutional neural networks (CNNs) has achieved impressive results in the field of image classification. This paper is concerned with a new approach to the development of plant disease recognition model, based on leaf image classification, by the use of deep convolutional networks. Novel way of training and the methodology used facilitate a quick and easy system implementation in practice. The developed model is able to recognize three types of diseases of one plant and pesticides and/or fertilizers are advised according to the severity of the diseases. The type of green leaf diseases is recognized by CNN and KNN algorithms. After recognition, the predictive remedy is suggested that can help agriculture related people and organizations to take appropriate actions against these diseases.

Birds are the warm-blooded vertebrates constituting of class Aves, there are nearly 10 thousand living species of birds in the world with multifarious characteristics and appearances. Bird watching is often considered to be an interesting hobby by human beings in the natural environment. The human knowledge over the species isn’t enough to identify a species of bird accurately, as it requires lot of expertise in the field of Ornithology. This paper presents an automated model based on the deep neural networks which automatically identifies the species of a bird given as the test data set. The model was trained and tested for 20 species of birds with the total images 7637 and 1853 images for train and test respectively and the model has shown a promising accuracy of 98% when tested with the test datasets.

Deep Learning (DL) algorithms enabled computational models consist of multiple processing layers that represent data with multiple levels of abstraction. In recent years, usage of deep learning is rapidly proliferating in almost every domain, especially in medical image processing, medical image analysis, and bioinformatics. Consequently, deep learning has dramatically changed and improved the means of recognition, prediction, and diagnosis effectively in numerous areas of healthcare such as pathology, brain tumor, lung cancer, abdomen, cardiac, and retina. Considering the wide range of applications of deep learning, the objective of this article is to review major deep learning concepts pertinent to brain tumor analysis (e.g., segmentation, classification, prediction, evaluation.). A review conducted by summarizing a large number of scientific contributions to the field (i.e., deep learning in brain tumor analysis) is presented in this study. A coherent taxonomy of research landscape from the literature has also been mapped, and the major aspects of this emerging field have been discussed and analyzed. A critical discussion section to show the limitations of deep learning techniques has been included at the end to elaborate open research challenges and directions for future work in this emergent area.

In the very recent years, development of music recommendation system has been a more heated problem due to a higher level of digital songs consumption and the advancement of machine learning techniques. Some traditional approaches such as collaborator filtering, has been widely used in recommendation systems, have helped music recommendation system to give music listeners a quick access to the music. However, collaborative filtering or model based algorithm have limitations in giving a better result with the ignorance of combination factor of lyrics and genre. In our paper, we will propose an improved algorithm based on deep neural network on measure similarity between different songs. The proposed method will make it possible that it could make recommendations in a large system to make comparisons by “understand” the content of songs. In this paper, we propose an end-end model, which is based on recurrent neural network to predict user’s next most possible song by similarity. We will make experiments and evaluations based on Million Song Dataset and demonstrate how it outperformed the traditional methods.