Arya Aftab's Projects
Implementation of Denoising Diffusion Probabilistic Model in Pytorch
An implementation of DropConnect Layer (Dense, Conv2D, and Wrapper(for all TensorFlow Layers)) in Tensorflow 2
Light-SERNet: A lightweight fully convolutional neural network for speech emotion recognition
Official repository of "A Machine Learning Framework for Predicting Entrapment Efficiency in Niosomal Particles".
My Curriculum Vitae
Physics-based neural network with Sine Activation Function
A real-time application of the LIGHT-SERNET model
Implementation of Rotary Embeddings, from the Roformer paper, in Tensorflow
File Transfer Using Serial Protocol and Python
An Implementation of SincNet using Tenorflow 2.x.
An implementation of Sparse Layers in TensorFlow 2. x.
A simple code to synchronize subtitles
A repository containing the implementation of the SVM by TensorFlow 2.x
Prebuilt binary for TensorFlowLite's standalone installer. For RaspberryPi. A very lightweight installer. I provide a FlexDelegate, MediaPipe Custom OP and XNNPACK enabled binary.
Tumor type classification with traditional feature extractions and classifiers.
Active contours, or snakes, are widely used in medical image processing applications, mainly to locate the desired area boundaries. Gradient vector flow (GVF) field, like other methods of calculating external force fields, is proposed to address ordinary snake modelsβ problems, such as poor convergence in indentations and low accuracy in segmentation of objects owned weak borders. These problems are most pronounced in high-noise images, such as ultrasound images. In order to solve the problems more, we utilized the generalized gradient vector flow snake model using minimal surface and two steps converging using both vector based normalization and component-based normalization with distinct controlling parameters on active contour. We adopt minimal surface function to address the problem of low segmentation accuracy in other conventional methods. We also use two steps converging using both vector-based and component-based normalization with distinct controlling parameters to improve the snake curve converging into long and thin indentations plus higher accuracy in noisy and meandering areas. The results obtained and compared with other methods show that the proposed active contour model not only can converge better into long and thin indentations, along with maintaining weak boundaries but also shows higher accuracy in segmentation of tortuous areas, especially in noisy images.