Browsing by Author "Hamza Iqbal"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Design and fabrication of electromechanical power attachment wheelchair
    (UMT Lahore, 2021-08) Hamza Iqbal; Mustafeez Ur Rehman; Rao Muneeb Ahmad
    According to worldwide researches, it is investigated that around 15% of the total world population is suffering from various form of disability. Around 65 million people in the world rely in wheelchair for their physical movements. Since the invention of wheelchairs in 1655, the users are left with no more than two basic options i.e. standard wheelchair or an electronic wheelchair, but the problem to be addressed is whether a single unit or device can be designed to make use of two different devices. A specially designed power system for ordinary wheelchair solves above struggles of wheelchair users including speed, effort based movement and one dimensional approach towards travelling. Electromechanical Power Attachment is specifically designed to provide day-to-day wheelchair users with a relatively low cost device as compared to an electric wheelchair and a device which would act as a power system once attached with their wheelchair and as a result, would provide the users with more speed than an ordinary electric wheelchair, 6 to 12 mph as compared to 5 mph of electric wheelchair. In essence, power system and wheelchair can be attached and detached with ease by the user.
  • Thumbnail Image
    Item
    Roman urdu hate speech detection using deep neural networks
    (UMT Lahore, 2022) Abu Hurera; Hamza Iqbal; Muhammad Talha
    Lately we have seen enormous growth in numbers of social media users and it has turned into a necessary part of our lives but it has alot of drawbacks, amongst them one of the biggest is the problem of hate speech and hostile language which adversely affect society. Detecting offensive language on social media has become one of the major challenges. Many types of research have been done in this regard and most of the research work is done in major languages like English but there are still some languages where there is a lot of work that can be done. Roman Urdu is one of those languages because it is used by a big chunk of the audience in India and Pakistan. We all know how spewing hate or running a propaganda these days is so easy on social media. A Fake news can destroy your whole reputation. Keeping this in mind, Hate speech could be directed towards any race, religion, politics or about genders.Keeping this in mind we took a dataset of roman Urdu tweets with labels offensive and non-offensive. We are using word embeddings trained using word2vec by using around 7 million Urdu tweets. The next phase is working on preprocessing. After preprocessing is done, to classify its text modern deep learning models are applied. Some famous Sequential models are used for this purpose that are RNN(Recurrent Neural Network), Bi-LSTM and Bi-GRU.In Multiclass we faced two main problems: one was lack of data and other was imbalance data. Due to lack of data we didnt use Deep learning methods and used machine learning methods instead.To balance the classes in data we used smote. We have used TF-idf for vectorization. so we performed multiclass classification using Random forest,Multinomial Naive Bayes and XG-Boost.
  • No Thumbnail Available
    Item
    Surgical strike
    (UMT, Lhr, 2018) Moeed Ahmad; Hamza Iqbal
    This is the time of technology and world is doing more n more developments day by day in the field of technology. We also tried our best to take a part in this development to give a product with latest and innovative features. Our product is a game. This is much more than just a software. We will try to make it more exciting and enjoyable by its animations, script and graphics.