Qiankai Tong
@Icarus-K-LaplaceComputer Vision, Image Processing, Applied Math, AI, World Model, ill-posed inverse problem
Language Breakdown
Lines of code distribution across 15 owned repositories
I-Shaped Developer
I-shapedSpecialist — deep expertise in Python
Collaboration Network
Global Impact visualization
Repos
15
PRs
0
Growth
+18%
Top Collaborators
No collaborator data yet.
Coding Streak
Contribution activity over the past year
Top Repositories
A research-grade denoising framework utilizing Fractional Calculus and Structure Tensors to preserve fine textures in medical and scientific imaging. Features frequency-domain fractional filtering, multi-scale structural analysis, and Python/Numba implementation. GPL-3.0 licensed; ideal for MRI, CT, and microscopy restoration.
OptiVerse - Interactive Optimization Playground An interactive web application for visualizing and comparing 8+ optimization algorithms (Gradient Descent, Momentum, Newton, BFGS, L-BFGS, Conjugate Gradient, Trust Region) on 6+ test functions (Rosenbrock, Rastrigin, Himmelblau, etc.). Features 2D/3D visualization
Zero-FLOPs operating-point calibration for lightweight YOLOv8 on metal surface defect detection. Five-seed evaluation on merged NEU-DET and GC10-DET, comparing baseline, P2, and CLAHE variants. Includes training, threshold sweep, statistics, and reproducible scripts for industrial inspection.
Ultrasound enhancement lab for CAMUS NII/GT: log-domain speckle suppression, structure-aware detail preservation, and automatic grid search to balance CNR, ENL, and edge ratio. Includes reproducible evaluation, visualization panels, best-parameter export, and deployment-friendly modular Python code.
Flagship restoration framework fusing Meta-Learning with Fractional Calculus priors. Features a lightweight CNN for adaptive parameter prediction, Numba-accelerated iterative solver, and hybrid Quality/Speed execution modes. Designed for extreme noise conditions in scientific imaging (Thermal/Astro/Microscopy). GPL-3.0 licensed.
High-performance Python framework for removing cosmic rays and sensor stripes from wide-field astronomical images (e.g., Antarctic AST3-2). Features Laplacian-guided detection, photometric preservation (<0.2% flux loss), and JIT-accelerated processing. GPL-3.0 licensed; designed for scientific data pipelines.
High-performance, JIT-compiled denoising engine for uncooled thermal imaging. Features 60+ FPS processing via Numba/SIMD, native 16-bit RAW support, and robust dead pixel correction. GPL-3.0 licensed; designed for industrial inspection, thermography, and embedded edge deployment.
ALSD-Framework: A white-box image denoising system based on robust statistics and Wiener filtering. It features automated noise estimation (MAD), structure-aware feature extraction, and local SNR-guided fusion to balance noise suppression and detail preservation without deep learning or manual tuning.
Oracle-Salt-Pepper-Restoration: An engineering upper-bound baseline for impulse noise removal. By assuming perfect noise detection (oracle), this project implements a two-stage sparse inpainting strategy to establish the theoretical performance limit for blind restoration algorithms.
Scene-Adaptive-Denoiser: A unified framework for infrared impulse noise removal. It replaces binary detection with a continuous, saliency-guided soft fusion strategy. By integrating thermal target priors, structure consistency, and global scene modulation, it robustly protects hot targets while cleaning background noise.
Open Source Impact
Contributions to external projects
No external contributions found.