AI-ContentLab Skip to main content


New Posts

Introduction to CycleGAN

  In this article, we discuss the CycleGAN architecture. Here we discuss the CycleGAN architecture and explain how each architectural component can be implemented. Find the whole articles of this series on my Medium Profile : CycleGAN Series ¨ Introduction In this series of articles, we’ll present a Mobile Image-to-Image Translation system based on a Cycle-Consistent Adversarial Networks (CycleGAN) . We’ll build a CycleGAN that can perform unpaired image-to-image translation, as well as show you some entertaining yet academically deep examples. We’ll also discuss how such a trained network, built with TensorFlow and Keras, can be converted to TensorFlow Lite and used as an app on mobile devices. We assume that you are familiar with the concepts of Deep Learning, as well as with Jupyter Notebooks and TensorFlow. You are welcome to download the project code. In the previous article of this series , we discussed the concepts of conditional generative adversarial networks (CGAN). In this
Recent posts

Useful Medical Images Datasets For ML Projects

  Medical image datasets. Source Artificial intelligence (AI) has made impressive strides in healthcare, but one major challenge remains: lack of data. Deep learning algorithms need lots of data to be effective, and medical images are expensive and difficult to obtain due to ethical and resource constraints. This makes it hard for researchers outside the medical field to develop new AI tools. This story aims to help by providing a comprehensive list of medical image datasets to support deep learning research. The datasets cover various body areas and are categorized for easy reference. Another list of medical datasets was recently published: A List of the available medical large language models: Med-LLMs Medical Image Segmentation Types and Applications Health and scientific research MedPix : A free online collection of over 59,000 medical images from various patients. The Cancer Imaging Archive (TCIA) : A public, de-identified cancer images (MRI, CT scans, etc.) organized by disease

CIFAR-10 Classification Using Mixture of Experts

Mixture of Experts We all recently heard that ChatGPT and GPT-3 were made based on an approach called Mixture of Experts (MoE) . Such an approach has gained traction is the machine learning field which is a powerful paradigm that excels in handling complex, high-dimensional data. In this blog post, we embark on an enlightening step-by-step tutorial to develop, train, test, and validate a Mixture of Experts for the classification of images from the CIFAR-10 dataset .  To implement MoE for image classification , we leverage the CIFAR-10 dataset, a benchmark in computer vision. With 60,000 32x32 color images across 10 classes, CIFAR-10 is a challenging playground to showcase the capabilities of MoE. By the end of this story, you will understand the basics of a Mixture of Experts, and how to develop a MoE for basic and simple classification problems.  P.S. This is not a very theoretical article. it is rather a How-To article on getting started with MoE for classification.  Understanding M

Finetuning VideoMAE

  VideoMAE is a self-supervised video pre-training method that uses masked autoencoders to learn data-efficient video representations. The method is based on video masking with a high ratio, which improves the performance of video reconstruction and the generalization of video representations on small datasets. The authors of the paper show that VideoMAE is a data-efficient learner for self-supervised video pre-training, and that it can achieve impressive results on very small datasets without using any extra data. The code for VideoMAE is available on GitHub . VideoMAE architecture What is a Masked Encoder A masked autoencoder is a type of neural network that can learn to extract and map meaningful latent representations into high-dimensional space from data by training on large datasets of input samples. The method is based on masking random patches of the input image and reconstructing the missing pixels. It is based on two core designs: an asymmetric encoder-decoder architecture,

Knee Osteoarthritis Diagnosis: A Deep Learning Approach with Mixture of Experts

  Osteoarthritis (OA) of the knee is a degenerative condition affecting three knee compartments (lateral, medial, and patella-femoral), typically developing gradually over a span of 10 to 15 years [1,2]. Primarily caused by wear, tear, and progressive loss of articular cartilage, it can also result from infections leading to joint cavity damage, resulting in discomforts such as limited mobility, joint pain, and swelling [3]. Cartilage tissue alterations and damage are common in all joints, with the knee and hip joints being particularly susceptible due to their weight-bearing nature. Knee OA predominantly occurs in individuals aged over 55, with a higher prevalence among those over 65, and it is estimated that by 2050, 130 million individuals globally will be affected. Early detection and treatment are crucial for mitigating the progression of knee OA and enhancing individuals' quality of life [5]. The complexity of diagnosing and treating knee OA lies in its multifaceted nature,

You may like