Papers
15
Total Citations
3,166
H-Index
14
About
Colin Lea is a computer vision and machine learning researcher whose work has significantly advanced the field of temporal action segmentation and fine-grained activity recognition. Best known for his landmark 2017 paper "Temporal Convolutional Networks for Action Segmentation and Detection," which has accumulated over 2,000 citations, Lea pioneered a hierarchical deep learning framework that elegantly captures both local and global temporal patterns in video — setting a new standard for how researchers approach video understanding tasks. His research spans robotics, surgical data analysis, and human-robot interaction, reflecting a rare breadth of applied impact. His contributions to surgical AI are particularly notable: his 2017 benchmark dataset for gesture segmentation in robotic surgery (288 citations) provided the community with standardized tools to evaluate automated skill assessment, accelerating progress across the field. Earlier work on segmental spatiotemporal CNNs and convolutional action primitives further established him as a foundational voice in fine-grained recognition. Beyond pure research, Lea has demonstrated a commitment to real-world applicability, exploring end-user robot instruction for manufacturing and unsupervised trajectory segmentation for robot learning. His cumulative body of work has shaped how modern systems perceive, interpret, and respond to human activity.
Research Focus
Key Achievements
Top Papers
- 1Temporal Convolutional Networks for Action Segmentation and Detection2,028 citations · 2017
- 2
- 3Segmental Spatiotemporal CNNs for Fine-Grained Action Segmentation248 citations · 2016
- 4Learning convolutional action primitives for fine-grained action recognition116 citations · 2016
- 5A framework for end-user instruction of a robot assistant for manufacturing99 citations · 2015
- 6
- 7
- 8
- 9System events: readily accessible features for surgical phase detection35 citations · 2016
- 10