Sachin Shah

github linkedin

Computer Vision and Machine Learning Projects

[Paper] TiDy-PSFs: Computational Imaging with Time-Averaged Dynamic Point-Spread-Functions
Proved that time-averaged PSFs generated by dynamic phase masks are fundamentally more expressive than PSFs from a static phase mask. Demonstrated substantially improved performance on monocular depth estimation and extended depth-of-field imaging. Published in the International Conference of Computer Vision (ICCV) 2023 Proceedings. [paper] [arxiv]
[Paper] Discovering an Interpretable Latent Space for Measured BRDF
Research in simultaneous learning of measured BRDF compression and synthetic material generation. Utilized beta-variational autoencoders to disentangle the learned latent space. Published in Computer Graphics International 2023's Lecture Notes for Computer Science. [arxiv]
[Poster] Face De-occlusion Research
Introduced expression-aware loss function to learn better face reconstructions to improve facial expression recognition accuracy. [poster]
Image Segmentation
Comparison of classical image segmentation techniques: K-Means, MeanShift, Split & Merge, and Histogram Thresholding. [demo] [code]

Education Projects

[Paper] WAVE: A Web-Based Platform for Delivering Knowledge-Driven Virtual Experiences
Research and development of automated virtual avatar delivery through the web. Used in support of NSF and Department of Education grants. Published in IEEE Computer Graphics and Applications. [paper] [wave]
[Paper] RAISE: Robotics & AI to improve STEM and social skills for elementary school students
Research in improving social skills of students with disabilities through teaching basic coding and virtual AI companions. Funded by the Department of Education. Published in Frontiers in VR. [paper]
Developed session observation tool to tag and analyze events. Supports live tagging and multi-session analysis. Used as a coaching tool to train teachers and label videos. [tool]

Statistics in Debate

[Paper] Quantifying Skill in Lincoln-Douglas Debate
Research and evaluation of ranking systems such as Elo and Glicko-2 on Lincoln-Douglas Debate rounds. Successfully predicted the 2021 Tournament of Champions winner. [article]
[Paper] Statistical Analysis of the Impact of the Transition to Online Tournaments in LD Debate
Determine affect of online debate on participation, speaker points, and side bias. Notably discoveries on pool size changes and speaker point inflation. [article]
[Paper] Statistical Analysis of Side-Bias in High School Lincoln-Douglas Debate Article Series
Web scraper to pull 40k+ round data points to determine bias. Results now cited as evidence in national debate rounds. [february 2021] [code]

Robotics and Autonomous Control Systems

[Talk] Custom Vision for FTC, 2019 FIRST Championship Robotics Conference
Presented research on custom vision and machine learning for FTC teams. Discussed classical image processing, optimization methods, and deep learning approaches. Invited back for the 2020 World Championship. [code] [slides]
Lunar Knights Robotics Team Control Systems Lead
Compete in NASA Robotic Mining Competition. Lead design discussions and teach control system workshops. Wrote custom C++ and Python libraries for control systems. [code] [notes]
FIRST Robotics FTC Team Control Systems Lead
Competed in FIRST Tech Challenge (2015 - 2019). Inspire Award Winner (2018 World Championship) - top overall team. Robot Semifinalist (2017 World Championship). Specialized in custom vision and machine learning algorithms. [code]


Code Golf
Compete in fewest bytes coding challenges. Top 1% golfer. [profile]
Sorting Methods
Graphical demonstration of common sorting algorithms (Bubble, Selection, Insertion, Merge, and Quick). [demo] [code]
Assorted Projects
Ongoing repository to solidify programming concepts. Projects included: interpreter for custom language syntax, steganography, and fractals. [code]