Projects
repaper
An open source python package to create an editable PDF form from a handwritten form.
Transfomers LayoutLMv3 Huggingface Pytorch EasyOCR PDF Google Forms
cricket songs classification
Fine-tuning large AST model for cricket songs classification using mixed precision training.
AST - Audio Spectrogram Transformer Fine-tuning Mixed precision training
mailcheck.ing
Full stack app and real time API to check email.
AWS FastAPI Docker Oracle Cloud Infrastructure
voice conversion transformer
Pretraining transformer seperating linguistic features and voice identity to achieve any to any voice conversion
Voice conversion Attention Transformer PPG BNF Speaker embeddings
mixrNet
Using mixup data augmentation technique as regularization and improving the ResNet50 architecture performance
Mixup Regularization ResNet50 Image Classification Pytorch
image colorization
Grey scale to RGB colorization using UNET architecture and VGG feature loss
unet VGG feature loss Lab space image CNN regression Pytorch
semantic segmentation - Thesis
Trained models on mitade20k dataset and finetuned models by class imbalance methods and Yolo-object detection method to remove false-positive intersections
Semantic segmentation Pytorch Segnet