Academic/Learning Projects

1. Image Captioning using Transformer

Built a Transformer decoder with custom positional encoding and attention layers, trained on 30,000 images to generate descriptive captions.

2. Self-Supervised Learning

Trained a ResNet18 using self-supervised rotation prediction, then fine-tuned for CIFAR-10 classification, achieving 80% accuracy.

3. Generative Adversarial Networks

Implemented DCGAN & LSGAN from scratch

4. Pixel-Level Semantic Segmentation

Utilized U-Net architecture to segment Corn & Soybeans from other crops on RapidEye Satellite Image.

5. Predictive Modeling

My first ever Machine Learning project.