It’s looking like 2016 will truly be the year of the drone. While drones have been used for recreational purposes and businesses are integrating drones into their business strategies, there are still technical, legal and environmental obstacles for drones to overcome. While many tech companies are interested in making drones better, many governments are still unaware of how drones work and how beneficial they could be to society.
But what if the environmental obstacles could be worked around? One report shows that environmental obstacles are a thing of the past for drones. MIT researchers have demonstrated a drone that can avoid trees while flying at high speeds in an open field. Swiss researchers are now showing that drones can navigate forest trails better than human beings using deep learning neural networks.
The researchers, a team gathered from the Dalle Molle Institute for Artificial Intelligence, the University of Zurich and NCCR Robotics, trained drones to navigate through trails in the Swiss Alps by mounting three GoPro cameras that provided forward, left and right facing views to a headset. The researchers got 20,000 images to develop and teach an algorithm to the drones that would allow them to successfully avoid hitting trees while in flight.
This allowed drones with only a single camera to navigate previously unexplored (to drones) trails without any human interaction whatsoever. “The algorithm was even better than humans at determining the correct direction of the trails it traveled on,” the researchers said. They claimed that an autonomous drone could predict the correct direction of a trail at 85 percent accuracy, compared to a human being’s 82 percent accuracy.
“Previous literature focused on trail segmentation, and used low-level features such as image saliency or appearance contrast; we propose a different approach based on a Deep Neural Network used as a supervised image classifier. By operating on the whole image at once, our system outputs the main direction of the trail compared to the viewing direction. Qualitative and quantitative results computed on a large real-world dataset (which we provide for download) show that our approach outperforms alternatives and yields an accuracy comparable to the accuracy of humans that are tested on the same image classification task,” the study’s abstract notes.
Thanks to the initial study’s success, the researchers intend to use these drones for search and rescue operations in remote areas, which could assist in the search for survivors after natural disasters such as forest fires and earthquakes. But does this mean that machine has won its battle over man? Not quite. The researchers said that the results are in preliminary stages and that further research is required. However, it signals a development in the capabilities of machines equipped with deep learning neural networks, and serves as an example of slight mechanical superiority over humans.