Mark Robinson
AI Research Scientist @ Kostas Research Institute (KRI)
MSc Robotics Engineering · Worcester Polytechnic Institute (2021)
BSc Computer Engineering · The University of Alabama (2020)
As a researcher at Kostas Research Institute, I develop and deploy advanced perception models, design semi-automated calibration interfaces for camera-LiDAR and cross-modal imaging systems, develop AR applications, among other projects. I'm motivated by my desire to learn, make the world a better place, and improve lives. When I am not writing code, you can find me playing basketball 🏀 or enjoying Downtown Boston or any of Cambridge's Squares
Model Zoo Exported Model Weights
Downloadable model artifacts
| Model Name | Task | Framework | Format | Input | Prec | Size | Metric | Link |
|---|---|---|---|---|---|---|---|---|
Depth Anyting V2 | Monocular Depth Estimation | Pytorch | ONNX | 1x3x518x518 | FP32 | 1300 MB | - | Download → |
DINOv3 Convnext Base pretrain lvd1689m | Image Embedding | Pytorch | ONNX | 1x3x224x224 | FP32 | 350.4 MB | - | Download → |
RTDETRv2 r18vd_120e coco rerun48 | Object Detection | Pytorch | ONNX | 1x3x640x640 | FP32 | 80.4 MB | - | Download → |
Research Projects
mss
Monocular Semantic SLAM pipeline utilizing TensorRT-accelerated Depth Anything v2, RTDETRv2, and DINOv3 models for depth estimation, object detection, and scene recognition/object tracking
View Project →wiys?
Google Cloud hosted perception ML platform. Utilizing WebRTC for browser-to-server communications, FastAPI for web architecture, and TensorRT compiled visual models for real-time performance. Containerized backend with dependencies and deploment recipes
View Project →taey
Real-time RGB-D SLAM pipeline integrating TensorRT-accelerated ViT-32 embeddings for keyframe feature extraction, FAISS-based visual similarity search, and GTSAM iSAM2 for incremental pose graph optimization
View Project →U-Net
Semantic segmentation model trained on Cityscapes.
View Project →Technical Skills
Programming
C++, Python
Frameworks
PyTorch, TensorRT, ROS 2, OpenCV, PCL, CGAL, GTSAM
Domains
Computer Vision / Perception, Camera Calibration, Sensor Fusion, SLAM, Embedded AI, Machine Learning, Robotics
Articles & Media
Repair Technology Exercise (REPTX) Distance Support
Image from the DVIDS article featuring the UNREP building exercise at NSWC PHD.
Read More →REPTX Distance Support: Remote Support Capability Validated
Highlighted the remote technology-assist capabilities validated during the exercise.
Read More →Structural Impact Assessment Scenario
Self Defense Test Ship damage assessment as part of REPTX 2022.
Read More →