Level Theta S images

My Virtual Forest project is still running strong and generates tons of spherical images (currently ~50GB). However, the post on which the camera sits is not perfectly level.  The Theta S camera normally compensates for this using an internal gyroscope which detects pitch and roll of the camera.  Yet, when downloading images directly from the camera no adjustments are made and the pitch and roll data is merely recorded in the EXIF data of the image.

As such I wrote a small bash script which rectifies (levels the horizon) in Theta S spherical images using this internal EXIF data. This is an alternative implementation to the THETA EXIF Library by Regen. I use his cute Lama test images for reference. All credit for the funky images go to Regen. Below is the quick install guide to using my script. I hope it helps speed up people’s Theta S workflow.


Download, fork or copy paste the script from my github repository to your machine and make it executable.

$ chmod +x theta_rectify.sh


$ theta_rectify.sh image.jpg

The above command will rectify the image.jpg file and output a new file called image_rectified.jpg.

Visual comparison between my results and those of Regen’s python script show good correspondence.


The script depends on a running copy of exiftools, imagemagick and POVRay. These tools are commonly available in most Linux distros, and can be installed on OSX using tools such as homebrew. I lack a MS Windows system, but the script should be easily adjusted to cover similar functionality.

On reviewing, publons and privacy

Recently I came across Publons as a way to get credit for your reviewing efforts. At first I was rather intrigued. It does sound like a good idea as I often bemoan the burden which is reviewing at times, and the little reward it brings (often because of the quality of the work). I was even more intrigued as five years ago I was runner-up in Elsevier’s peer-review challenge, trying to resolve the ailing peer-review system. The winner suggested something along the lines of Publons.

However, Publons or any badge system want to make you believe that your review is worth something outside it’s academic context. Yet, it does not contribute to a workable solution for the core problem of the academic peer-review process which is ease-of-use and the quality of the review. It perverts the peer-review system with false incentives and a race to the bottom if poorly executed. Publons claims for publishers state a decrease in review times and accepted reviews. This suggests that Publons change incentives to accept reviews, and potentially the number of accepted manuscripts. Assuming that time is a limited resource for most scientists, increasing the number of reviews should decrease the time spend on them, letting errors slip through.

Even less surprising is that given that this is a for profit venture and they do not go lightly when it comes to privacy. Checking the privacy statement (below) basically states that all the data you submit (full reviews if possible) can be used for data mining, and reselling to advertisers or publishers alike.

In short, Publons is a niche data broker, contrary to sweeping approach Google and Facebook use. Added value is generated in the form of a virtual badge, with little or no real world value, providing only an extra account to track and performance anxiety that goes with it and the privacy you sign away. The badges potentially shift the reviewer acceptance rates due to time restrictions and moral hazard. We should not be speeding up science, we should be increasing rigour, reproducibility and quality.




CitSci Jungle Rhythms project finished!

After an amazingly short ~475 days the Jungle Rhythms citizen science project finished !

I hereby would like to thank all citizen scientists who contributed to the project and made it into a success, transcribing ~30K site years of tropical phenology data in no time.

I will now start the post-processing. I will report on this shortly as I make my way through the large dataset consisting of more than 300K classifications and additional remarks collected from the Talk forums. Intermediate results have shown that the data collected are of an outstanding quality, illustrating the swift and accurate work by citizen scientists. I’m confident the complete dataset will be of an equal high quality.

Once more, my thanks and gratitude go out to all volunteers who made this possible !

Robots are -also- after your academic job

Rapid development in machine learning and artificial intelligence has allowed previously complex consumer applications and jobs to be automated at an ever increasing pace. But make no mistake, machine learning will also displace academic jobs with only those remaining which wield the right tools combined with the best and brightest ideas.

Voice assistance allow you to order anything online, translation services process natural language ever better, ordered packages are picked by robots, and soon a drone might deliver them. All these advances in automation are due to recent software and hardware developments which result in job loss.

Yet, in academia the rise of automation and ‘robots’ stirs less of a concern. The general feeling remains that core functions of academic jobs are too complex to be sufficiently automated. This core function is the creation of new hypothesis to be tested. However, a large part of academic work, especially in ecology, still relies on encyclopedic knowledge, collection management and often tedious work. Encyclopedic knowledge and many of these mundane academic tasks can be reduced to harder classification or sorting problems (of samples stored in collections and data sets).

As with the soon to be obsolete Uber driver, harder classification problems,  such as driving a car, become easier to solve using advanced machine learning techniques. One can argue that automation will free up time to do more hypothesis testing, and increases the pace of science. This arrival of machine learning into mainstream research can be seen in a number of recent high profile publications from classifying leaves in paleobotany to mapping poverty.

I argue that in the near future it will be a requirement to wield the tools to implement (large scale) machine learning approaches to remain competitive in most academic fields. With austerity and budget cuts, a failure to do so might lead to some academics being replaced by evermore sophisticated algorithms and not necessarily lead to a shift in their job content. A urgent shift is needed in the skill sets of many scientists, extending past statistics into the fields of computer science and machine learning.


Citizen Science restores trust in science; isn’t exploitative

A recent review comment made the claim that Citizen Science (referring to my Jungle Rhythms project) is exploitative. With due diligence on part of researchers, not only is this comment misguided, it also is a testament to a pervasive ivory tower way of thinking about science.

Science is often perceived as a field of the select few with limited interaction between the scientific world and the public. Or as aptly put in the Guardian Science section: “Science is the invisible profession. Most people have no idea what scientists do, and may harbour a vague feeling of suspicion or uneasiness about the whole endeavour.”

This lack of transparency has been abused many times over to create doubt and confusion in order to push a political agenda, fueling among others climate skepticism. In addition, a lack in transparency and limited communications creates a less educated public and one which is less used to dealing with complexity.

Citizen science provides a way to counter all these issues. It allows citizens to actively contribute to science, directly communicate with scientists and at times attain PhD worthy knowledge through self-study. In today’s society with increasing distrust in science through fake news and “alternative” facts this direct and transparent communication about science between scientists or science communicators and the public is key to retain or restore trust in science.


© 2018. All rights reserved.

Powered by Hydejack v7.5.1