Ben Upcroft

About

Ben Upcroft is a pioneering researcher at the intersection of deep learning, computer vision, and agricultural robotics, whose work has fundamentally advanced the application of artificial intelligence to autonomous robotic systems. Best known for the landmark paper "DeepFruits" (2016, over 1,079 citations), he demonstrated that deep convolutional neural networks could achieve accurate, real-time fruit detection — a breakthrough enabling practical autonomous harvesting platforms. His broader contributions span visual place recognition, where his investigations into ConvNet features for robotic navigation and SLAM (532 citations) helped bridge the gap between computer vision research and real-world robotics deployment. Upcroft has also made significant strides in vision-based reinforcement learning for robotic motion control and semantic mapping on mobile robots, showing how robots can learn and navigate complex environments without environment-specific training. His work on lightweight deep neural network architectures for agricultural robots and vision-based obstacle avoidance reflects a consistent commitment to making AI practically deployable on resource-constrained platforms. Further contributions to crop harvesting — including sweet pepper pose detection and occluded crop recognition — underscore his focus on solving tangible agricultural challenges. Collectively, his research portfolio has garnered over 2,700 citations, establishing him as a leading figure in intelligent agricultural robotics.

Research Focus

Key Achievements

Top Papers

1. 1
   DeepFruits: A Fruit Detection System Using Deep Neural Networks

   1,079 citations · 2016
2. 2
   On the performance of ConvNet features for place recognition

   532 citations · 2015
3. 3
   Towards Vision-Based Deep Reinforcement Learning for Robotic Motion Control

   215 citations · 2015
4. 4
   Mixtures of Lightweight Deep Convolutional Neural Networks: Applied to Agricultural Robotics

   179 citations · 2017
5. 5
   Vision‐based Obstacle Detection and Navigation for an Agricultural Robot

   152 citations · 2016
6. 6
   Place categorization and semantic mapping on a mobile robot

   146 citations · 2016
7. 7
   On the Performance of ConvNet Features for Place Recognition

   135 citations · 2015
8. 8
   Sweet pepper pose detection and grasping for automated crop harvesting

   91 citations · 2016
9. 9
   The ACRV picking benchmark: A robotic shelf picking benchmark to foster reproducible research

   81 citations · 2017
10. 10
    Visual detection of occluded crop: For automated harvesting

    80 citations · 2016

Key Collaborators

Contact & Links