Smart Vision update

Noah and friends have done it! The Open CV (computer vision) is running on a Raspberry Pi. To ensure that the system performs well in real-time they had to “overclock” the system and to add a heat-sink to prevent the chip from burning. The result is a fast, low-cost and low-energy smart camera that can be used for wildlife census and anti-poaching missions.

The recognised ‘objects’, in our case, are humans, elephants, tigers, and other species. The outcome is communicated with Cluey to inform park rangers in real-time. The data may also be used by census-researchers. In that case the classified images may be collected periodically.

Below you see the sneak-preview of the cloud-based smart-cam training console. This console may be used by experts or the public to improve our classifier (for the experts: we use a mix of classic learning, machine learning, and deep learning). Once the detection accuracy of the smart-cam is sufficient, the sensor can be placed in the field. To reduce communication cost, only the class will be send. To increase confidence in the system, a thumbnail of the recognised object may be send as well.

sneakpreview smart cam.png

Can’t wait to experiment with the smart cam? Feel free to contact us to discuss how we can speed up time to the field!

DIY Smart Computer Vision

The problem

If you have a camera to detect burglars, an alarm system to detect opening doors, and a smoke detector to alert you in case of fire, wouldn’t it be nice if all these systems could be presented to you through one simple app? If you already have an integrated system like that, you probably bought all of the components at the same supplier. Adding third-party or Do-It-Yourself (DIY) sensors, or combining the data with other data sources, such as weather stations or the GPS-position of your mobile phone, is probably hard or impossible. Options to combine the data, however, would enable you to make your systems really smart…

The solution

To allow you to connect any type of sensor you wish, we are developing an open source Application Programming Interface (API). Any hobbyist or programmer can use it to connect his of her own device to the SCCSS-sensing platform. From there, the data may be presented in Cluey or WildCAT, or exported to CSV, XML or JSON, for further analysis.

Low-cost Smart Computer Vision Camera

A first trial project has been adopted by 4 students of Technasium Keizer Karel College Amsterdam. Noah, Robin, Celio and Dimme have the ambition to turn a standard webcam into a low-cost Smart Computer Vision Camera. That’s a camera which does not just take pictures, but which can be trained to takes pictures only when a person walks along, or a dog, tiger, elephant, or whatever else it has been trained for. Once the object of interest has been detected, a small image is created and sent to the Cluey-app.

A tough task, involving the mastering of Open Source Computer Vision software on a Raspberry Pi, Python programming, a little engineering, and lots of endurance, fun and enthusiasm!

Noah, Robin, Celio and Dimme expect to finish the project this summer. They will publish the source code and the “How to” in GitHub, thus making it available to the public for free.

The first results are looking good!


Screen print: Noah captured by the team’s engineering masterpiece, the low cost Smart Computer Vision Camera!

Sensing Clues at TEDx!

You should listen to the TEDx-talk I held at TEDxRotterdam last November,

  • if you feel angered about elephants being killed for their ivory,
  • if you want to be challenged to do something about it – by doing what you know best!

Looking forward hearing from you!

Jan-Kees Schakel, CEO Sensing Clues

ps please share the story with your friends!

Sensing Clues goes to Nepal!

After a summer of hardcore development, Sensing Clues is ready for the next step. Together with the Himalayan Tiger Foundation we’ll travel to Nepal to examine and demonstrate the potential of our sensor and real-time intelligence tools. Think of:

  • passage detection,
  • lingering detection,
  • group size detection,
  • first-time seen detection,
  • light-beam detection,

and many other profiles which can be used to inform rangers of alarming occurrences in protected areas. Following is a short summary of what we did this summer.

The Trespasser, a sensor designed to detect electronic devices, has reached it next level of maturity. In a pilot study conducted with the Dutch National Forestry Department we were able to detect the difference between hikers passing a nest of a protected hawk and people lingering near the nest. Based on the number of devices we could estimate the number of persons at the scene. This is an important feat, as “big-5 poachers” often work in groups of 3 to 4 persons. Recognizing the number of people lingering near a waterhole or other critical spots thus gives rangers an early warning of a threatening situation.

Our new Serval-soundscape sensor has passed its first milestones. Data acquisition, a tech-word for recording and storing the sounds in a ready-to-process format, is ready. So is the store-in-memory function that ensures that sounds do not get wasted when connectivity is lost for a while. The next step is to incorporate the recognition algorithms and to establish connectivity. Both are within reach. As soon as time and funds permit we will start field-testing.

Power supply is a critical issue when working in remote areas. As there is no off-the-shelf solar solution that meets our tough outdoor requirements, which includes proper camouflaging to avoid detection by poachers, we needed to develop a tailor-made solution. The good news is: we did. The first results are promising. In two days time the solar panel is able to load a car battery that keeps the sensor alive and kicking for over 6 weeks. Hence, protection operations of 3 to 6 months have become within reach.

In the mean time we experimented with the setup of a LoRa-network. Such networks are comparatively cheap and can be deployed where cellphone coverage is lacking. As with all technology, the road to full-scale use is bumpy. If not properly configured, reaching a proper range is troublesome. Knowing the problem is half the solution. So we are now working on the second half.

All that being said, most of our time was spent on the development of Cluey, our fast-response coordination app, and its backend, which in fact constitutes an affordable sensing-and real-time analytics platform and intelligence tools. Our platform constitutes a dozen servers, software packages, has very high security standards, and is maintained by our engineers. Regularly, such systems are prohibitively expensive for a single park or NGO. By offering the platform as a service, however, we bring it within reach of even the most modest NGO.

Ps. local governmental organisations in the Netherlands have shown an interest in these tools also, which accelerates their development!

Our hawks have left their nest – mission completed!

Mission completed! This weekend the young hawks stretched their wings and made their first flight.

For almost two months the rangers in the Southern Netherlands have been on the alert for poachers, as the nest has been robbed for several successive years. Local bird lovers complained, but the problem was difficult to tackle. Attempts to catch the poachers through camera traps failed, as the cameras were stolen.

This year a Trespasser was hidden near the birds-nest to notify the rangers when people would come near. The sensor triggered 16 times. Based on a smart algorithm we are able to determine whether a detected person is bypassing or lingering at the scene. We were especially interested in the latter, which happened twice. The first time was on the first of June, when the young were still very small. Great was our relieve when it proved to be one of our rangers checking whether the birds were still safe. The second time foresters were busy marking trees near the nest.

Now the birds have left their nest on their own – the first time since years. We are so happy about it! A pity that we didn’t catch a poacher, but so be it. The sensor operated well and unattendedly for over 6 weeks. The rangers are enthusiastic about them as the sensors can distinguish between people and other moving targets and can be placed completely out of sight (see photo below – just try to find it 😉

Time to celebrate and move on to the next project!

camouflaged sensor

The clock is ticking…

About one week to go before the young of our hawks will fly out and discover the world! So far we managed to shield off the poachers, who have been stealing chicks for years.

It’s a project in which we learn while doing. As we knew, in fast-response organizing every second is of the essence. Easy to say. Quite complex to accomplish technically.

Today we scrutinized the process steps from signal detection to alert. A process which from sensor to end-user app covers over 15 time critical components.

First LoRa field-tests for the protection of endangered birds

Today we have set up our first LoRa-network for the Dutch National Forestry Departement. In the next two months we are testing the communication between the gateway, the outdoor field sensors, and our secure cloud services.

Concurrently we installed two Trespassers for the protection of endangered birds of prey. The last few years they have been robbed of their young by poachers. Together with the rangers of the State Forestry Department we aim at keeping them safe.

placing antenna
 Placing the LoRa gateway antenne

testing gateway
Testing connectivity with our secure cloud services

testing field connectivity
Testing connectivity in the forest