Fangda Li Email & Phone Number

California, United States

California, United States

Pleasanton, California, Us

West Lafayette, In, Us

• Designed a generative image-to-image translation framework that translates H&E-stained images into various IHC stains while accurately predicting the diagnosis-critical molecular representations. • By using a novel adaptive contrastive learning based objective, the training of the virtual IHC-restaining network is robust to the inevitable and often severe inconsistencies in groundtruth H&E-IHC image pairs.• Designed a generative stain augmentation network for augmenting H&E-stained cell images with synthesized yet realistic stains that can help desensitize downstream application-specific models to stain variations.• By disentangling representations for cell morphology and stain while using a Laplacian Pyramid based architecture, the model can achieve transformation to arbitrary stains with high efficiency.• Designed an end-to-end automated, real-time, machine learning-based semantic segmentation framework for automatic explosive recognition in 3D dual-energy X-ray CT images of airport passenger checked baggage.• By using an ensemble of deep learning and boosting algorithms, the framework achieved state-of-the-art detection rates while maintaining low false alarm over a large-scale dataset (5k+ real-world baggage scans).• Developed a GPU-accelerated model-based CT image reconstruction algorithm for dual-energy X-ray CT that outperformed state-of-the-art approaches in both signal-to-noise ratio and convergence speed. • Contributed to installing and maintaining an OpenStack cloud computing framework for all research at RVL.• Developed a novel motion planning algorithm that leverages recursion and gradient descent to find efficient yet smooth trajectories for robot navigation in congested and narrow spaces. • Developed computer graphics software in OpenGL for 3D interactive apple tree pruning simulation.

West Lafayette, In, Us

Deep Learning, ECE60146, Spring 2023• Graduate level class on CNN, RNN, Transformer, GAN, etc.Computer Vision, ECE664, Fall 2022• Graduate level class on Geometric Computer Vision, e.g. Stereo Reconstruction.Digital Systems Senior Design, ECE477, 2019, 2021, 2022• Senior undergrad level class on Embedded System design and programming.

Pleasanton, California, Us

• Collected data and developed a framework for performant nuclear instance segmentation in H&E-stained histological images.• Designed and implemented generative adversarial networks for normalizing the wide range of variations among the H&E stain appearances.

Guangzhou, Cn

• Developed a true scale estimation module for monocular ORB-SLAM by integrating IMU inputs using Extended Kalman Filter on mobile devices.• Conducted literature review on and implemented various algorithms for the Multi-Armed Bandit problem.

Hannover, Lower Saxony, De

• Improved the Random Forest algorithm for unbalanced datasets by integrating class importance and leaf weights.• Proposed algorithm outperformed the state-of-the-art on real-world face detection and traffic sign recognition datasets.

Doctor Of Philosophy - Phd, Computer Engineering

Master Of Science (M.S.), Computer Engineering

Bachelor Of Science (B.S.), Electrical And Electronics Engineering