Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Self-supervised learning for Human Detection in Search and Rescue Missions
Use techniques from self-supervised learning to reduce the workload and eventually completely spare the human needed for dataset labeling in the context of a deep learning based human detection framework.
Keywords: Unsupervised learning, computer vision, object detection, human detection
ASL has been working on fixed-wing UAVs since 2007, more recently shifting the focus from long-endurance solar powered UAVs to intelligent autonomous drones. For search and rescue missions, ASL’s fixed wing team has developed a deep learning based human detection framework that uses the imagery from a thermal and RGB camera. To continuously improve the detection performance of the framework it is important to provide enough training data to the DL architecture. However, human labeling (bounding box at x,y) in datasets remains extremely cumbersome and therefore costly. To this end, the goal of this project is to use techniques from self-supervised learning to reduce the workload and eventually completely spare the human needed for dataset labeling.
Related Literature
- Kümmerle et al., “Real-Time Detection and Tracking of Multiple Humans from High Bird’s-Eye Views in the Visual and Infrared Spectrum”, 2016
- Wang et al., “Fast and Accurate, Convolutional Neural Network Based Approach for Object Detection from UAV”, 2018
ASL has been working on fixed-wing UAVs since 2007, more recently shifting the focus from long-endurance solar powered UAVs to intelligent autonomous drones. For search and rescue missions, ASL’s fixed wing team has developed a deep learning based human detection framework that uses the imagery from a thermal and RGB camera. To continuously improve the detection performance of the framework it is important to provide enough training data to the DL architecture. However, human labeling (bounding box at x,y) in datasets remains extremely cumbersome and therefore costly. To this end, the goal of this project is to use techniques from self-supervised learning to reduce the workload and eventually completely spare the human needed for dataset labeling.
Related Literature
- Kümmerle et al., “Real-Time Detection and Tracking of Multiple Humans from High Bird’s-Eye Views in the Visual and Infrared Spectrum”, 2016 - Wang et al., “Fast and Accurate, Convolutional Neural Network Based Approach for Object Detection from UAV”, 2018
- Literature review on self-supervised learning
- Implementation of the most promising approach
- Validation with (existing) labeled data
- Literature review on self-supervised learning - Implementation of the most promising approach - Validation with (existing) labeled data
- Courses and experience in computer vision and (deep) learning
- C++, python
- Courses and experience in computer vision and (deep) learning - C++, python
For more information, visit: https://docs.google.com/presentation/d/1DVAy-Jl4dDyL4uEeAYm6aD9b2YS_-oTauESIQ7nvhxE/edit?usp=sharing
Please send your CV and transcript of records to hitimo@ethz.ch
For more information, visit: https://docs.google.com/presentation/d/1DVAy-Jl4dDyL4uEeAYm6aD9b2YS_-oTauESIQ7nvhxE/edit?usp=sharing
Please send your CV and transcript of records to hitimo@ethz.ch