01
Computer VisionDeep LearningPython
Cataract Detection System
Real-time cataract detection from eye images. Built the full pipeline — model, inference API, and serving layer. 98.1% accuracy.
98.1% detection accuracy on real eye image datasets
Back to Projects
Deep LearningAudio MLNLP
Audio Captioning System
2023 · Live Project
The question was simple: can a model understand audio well enough to describe it in words? Built a captioning pipeline using VGG16 for feature extraction, CNN-LSTM for sequence modeling, and Transformer-based attention for context. The system generates contextual text captions from raw audio — no transcription, actual semantic understanding. More interesting than it sounds.
OUTCOME
Contextual captions generated from raw audio without transcription
TECH STACK
TensorFlowKerasCNN-LSTMTransformersVGG16
