Iqbal Singh

ViT Model Replication

• Reproduced Vision Transformer (ViT) from research paper, achieving 85% accuracy on CIFAR-10
• Enhanced model interpretability using attention visualization techniques

Image Classification with CNNs

• Built and trained a CNN model on a custom dataset, achieving 92% accuracy
• Implemented data augmentation and transfer learning for improved performance

Object Detection with YOLOv5

• Trained YOLOv5 for real-time object detection on a custom dataset
• Fine-tuned the model and optimized inference using ONNX for deployment

GANs for Image Generation

• Developed and trained a Generative Adversarial Network (GAN) to generate realistic images
• Applied techniques such as progressive growing and loss function optimization

NLP with Transformers

• Implemented a text classification model using Hugging Face’s Transformers library
• Fine-tuned a BERT model on a sentiment analysis dataset, achieving state-of-the-art results

Reinforcement Learning with DQN

• Built and trained a Deep Q-Network (DQN) agent for solving OpenAI Gym environments
• Optimized training efficiency using experience replay and target network stabilization