Publication

Projects

Work Experience

Deep Learning Research (Video Persod Re-id), Clarifai
(Guide-Dr. Michael Gormish)

• Implemented and mixed various techniques, resulting in 82.1% mAP accuracy where previous reported SOTA is 80.8% on MARS dataset for video-based Person Re-id task, accompanied by a detailed experimental study of model
• Built the Video back-end infrastructure for loading and training videos

Deep Learning Research (VAE-GAN for Audio), Rice University
(Guide-Prof. Anshumali Srivastava)

• Applied the VAE-GAN on audio spectrogram per frame to reconstruct phonemes of vocal recordings
• Compared English trained model with French counterpart to visualize phonetic differences, using T-SNE plots
• Attended High Performance Computing (HPC) BootCamp for parallel programming and working on GPUs