Deep learning animates water

Animating water isn’t easy.

Firstly, it’s a liquid so it can move multi-directionally. Secondly, it’s mainly transparent, so it needs to be clear (yet not entirely), and it refracts light. Thirdly it has a reflective surface which moves.

Because of the many different elements which need to be skilfully executed, and the significant computational processing requirements necessary to make a high quality simulation, animating water is a time-consuming and difficult challenge.

Recently however, researchers at Washington University have used deep learning to develop a system which shows water moving in a waterfall from just one still picture.

By using thousands of videos with fluid motion including seas, waterfalls and rivers, the process teaches a neural network to predict and animate how the moving water would appear from a single photograph. The neural network views the videos and then guesses the motion based on the first frame.

It was able to learn, based on the image’s context clues, what the motion was supposed to look like. Then, when its output was compared with the actual video, the network slowly learned what to expect from different states of flowing matter.

And the same method – which uses a seamlessly looping short video to give the impression of continuous movement – can be used to animate clouds, smoke, or any other material that flows.

The advantage of the method is that no extra information or user input is necessary. It simply needs the picture from which it uses the information gained by predicting the movement in the photograph when it was taken. This enables the system to determine the movement of each pixel and create the animation.

The team used the term ‘symmetric splatting’ to describe the movement of each pixel according to its predicted motion both past and future. Firstly, the researchers tried ‘splatting’ but that meant that as the pixels travelled down the waterfall, they disappeared from the top.

By predicting the past and future of the image, the predictions could be combined into one animation on a single loop which left no pixel gaps and allowed the endless movement of the animated image.

Because as yet the process has difficulty predicting the distorted appearance of an object under water, or the reflections on moving water, it works best for objects with predictable fluid motion. Nevertheless, it appears to have achieved a more realistic representation of moving water than many other software tools. 

Into the uncanny valley.

According to MacDorman & Ishiguro, 2006,

‘The uncanny valley can be described as the hypothesized relationship between the extent to which a humanoid entity resembles an actual human being and the emotional response such an entity evokes.’

Which actually means that when we experience humanoids on screen: 3D animations, virtual reality, photorealistic animation, robotics and lifelike dolls that are particularly realistic, we can find it inexplicably unsettling.

Somehow, observing a lifelike digital being that has been meticulously rendered can be simultaneously both disturbing and captivating. Though technology enables digital humans to appear virtually indistinguishable from an actual human, it’s the virtually indistinguishable part that causes the disquiet.

Because to many people, they actually look quite creepy, eerie, or even revolting.

Masahiro Mori first introduced the concept of the uncanny valley in the 1970s when he was a professor at the Tokyo Institute of Technology. He hypothesised that the more resemblance to an actual human a robot had, the more positive and empathetic was the emotional response from the observer – but only up to a certain point.

It appeared that at this point, the observer’s response becomes extremely negative, even to the point of revulsion.

However, as the robot’s appearance continues to become even more human, the levels of empathy and positive emotional response from the observer return – hence the analogy of moving first down then up the sides of a valley.

It is this disturbing in-between place which is known as the uncanny valley – the area in which our affinity with what we are viewing descends into feelings of unease or even fear.

But why do people have this response?

One explanation may be that we judge non-human objects by human standards which cannot be met. When the object is clearly non-human, we easily accept it as such but when it appears almost human, we focus more on its shortcomings.

In 2016, research suggested that that the further down in the uncanny valley a face is, the longer it takes to determine whether the face is human or not, eliciting a greater cognitive challenge. Prior to this, other research had suggested that it is this cognitive challenge which is associated with the negative emotions elicited.

If realism is the goal, and increasingly sophisticated technology enables the increasingly human-like appearance of virtual characters – which viewers are nevertheless still able to discern as virtual – then building a bridge over the uncanny valley is the challenge.

However, it could be argued that the eeriness and disquiet to be found in the uncanny valley is part of the appeal.

Animmersion Management Buyout

We are pleased to announce that on 16th February 2022 a management buyout was concluded for the Animmersion Group of companies lead by Sam Harrison and supported by its excellent management team and partners. 

We would like to take this opportunity to thank all the exiting investors, and founding partner Dominic Lusardi, for their tremendous support and help in growing the business, and wish them all the very best in their future activities. 

Animmersion has exciting plans for the future. Watch this space!

VR, AR and remote immersive collaboration

Remote and hybrid working looks like it’s here to stay in various configurations and for those who are home and office-based, it’s becoming more and more the norm.

However, for manufacturing, engineering and construction, and other industries which carry out complex projects from disparate locations and rely on 3D data, the transition is far more problematic.

Animmersion UK | Digital Visualisation | Virtual House

Inevitably, 2D screen versions cannot provide the same detail or enable full communication and consensus which, in turn, can delay the project costing time and money.

Potential solutions may lie with virtual and augmented reality (VR and AR) as disparate teams working with mission-critical 3D data in conjunction with BIM and CAD data are in a far better position to communicate effectively and reach a consensus.

However, developing the appropriate technology is key as when displaying data beyond around 250,000 polygons, the experience degrades and so it is essential to be able to develop solutions that provide the requisite detail.

Once implemented however, businesses are not challenged by screen sharing constraints as people can collaborate in a 3D meeting space no matter where their location.

This saves time and money and enables swifter understanding and resolution of road-blocks and difficulties in the project and allows project modelling to be visualised and evaluated in a range of detail in different contexts.

In addition, oversite and approval of progress in construction projects can be reviewed by the appropriate staff through access to detailed, up to date and even real-time data.

The bigger and more complex the project, the bigger the risk, so the use of VR, AR and other advanced technologies are being increasingly adopted to improve productivity all round.

VR and mental health

A study in New Zealand involving a collaboration between New Zealand’s Otago University Mental Health Clinical Research Unit, Auckland Institute of Studies, Otago Polytechnic Auckland campus, and Xian Jiaotong-Liverpool University is highlighting the potential of using virtual reality (VR) to address mental health issues.

Animmersion UK | Virtual Reality Developers | Man with headset

Recently published in the Journal of Medical Internet Research Mental Health, the report on studies taking place between 2017 and 2021 examined the use of VR as a treatment for anxiety. The findings suggested that in varied situations the treatment was successful enough for it to be considered for use as a clinical tool, providing patients with a tailored VR experience to help treat anxiety and depression.

Benefitting both patients and doctors is the ability to have a VR view of inside the human body. In this way, patients can enter, virtually, a specific reconstruction of their own body’s, pathology and anatomy, and thence be taken through their individual surgical plan. This leads to higher patient satisfaction as they have an improved understanding of their treatment.

Although it has been reported that a fear of needles (trypanophobia) could be a barrier to otherwise willing individuals becoming vaccinated against Covid-19, new software has been developed that uses virtual reality to distract patients so they can receive the injections.

The programme had been tested out by patients receiving flu jabs while wearing the VR headset. One patient said that he could ‘barely tell’ when the injection was taking place and that he would recommend the app to anyone who is afraid of needles.

Obviously, this technology has huge potential to help alleviate phobias or anxieties in other applications and settings such as social situations, spiders, fear of heights or flying.

The interdisciplinary nature of the research between artificial intelligence, virtual reality and mental health therapies in clinical use for personalised, tailored VR technology, combined with the diminishing expense VR headsets over time of should help increase the quality of treatments and improve outcomes.

Return to Real, Hybrid or Virtual Events?

Up until March 2020, live events were probably the optimal environment for product launches, exhibitions, and trade shows. Though costly and complex, the benefits of gathering many stakeholders in person, in terms of excitement, engagement and networking were universal.

Then the Covid pandemic hit, and the landscape changed. Cancellations and postponements were rife and there were massive disruptions to face to face interactions.

Businesses realised they must adapt quickly to maintain their market position by adopting new delivery models – going entirely digital or embracing hybrid digital and in-person solutions.

Now, however, venues have reopened, so does this call time on virtual events?

Not necessarily.

Many businesses acknowledge that even though they are now able to host in-person, they have found that virtual events can offer a unique experience, maximise attendance, boost brand recognition and attract a wider audience – particularly as some people simply will not be ready to return to attending live events.

In the near future at least, restrictions may come and go, but virtual and hybrid events keep people connected regardless. The option of digital attendance enables the maximum number of attendees to be reached while providing reassurance and thus improving staff engagement – while mitigating travel costs and neutralising the effects of bad weather.

Concerns that digital or hybrid events may not provide a full experience are misplaced as businesses recognise the unique benefits of remote solutions utilising 3D immersive technology. These include personalised virtual booths that can be integrated within a website or used for webinar engagement; virtual product showrooms which can be guided or free roam; virtual guided tours of facilities – such as house builds – which can be collaborative or individual.

Products and services can be communicated in an engaging and interactive way.

Importantly, digital and hybrid events can be integrated with a business’ existing marketing assets such as videos and product models and can remain available to view long after the event is over.

Altogether, a virtual/hybrid problem-solving approach can provide businesses with opportunities for excellent engagement, the ability to re-use exciting digital assets, maintain or increase market share, and reduce costs.

By utilising flexible and scalable visualisation technologies, businesses can future-proof their client experience, regardless of how rapidly the situation may change.