“Slow” marine animals show their secret life under high magnification. Corals and sponges build coral reefs and play crucial roles in the biosphere, yet we know almost nothing about their daily lives. These animals are actually very mobile creatures, however their motion is only detectable at different time scales compared to ours and requires time lapses to be seen.

Make sure you watch it on a large screen! You won’t be able to appreciate this clip or see individual cells moving in a sponge on a smartphone. This clip is displayed in Full HD, yet the source footage (or the whole clip), is available in UltraHD 4k resolution for media productions.

Visit my website to see more work: microworldsphotography.com
Learn more about what you see in this video: notes-from-dreamworlds.blogspot.com.au/2014/03/slow-life.html

The answer to a common question: yes, colors are “real” and not exaggerated by digital enhancement. I have only applied basic white balance correction. When photographers use white light on corals, they simply miss the vast majority of colors. Read more about fluorescence and why these corals are natural: notes-from-dreamworlds.blogspot.com.au/2013/06/fluorescent-colors-of-reef-coral.html

The duration of sequences varied from 20 minutes to 6+ hours.

=== Technical details ===

To make this little clip I took 150000 shots. Why so many? Because macro photography involves shallow depth of field. To extend it, I used focus stacking. Each frame of the video is actually a stack that consists of 3-12 shots where in-focus areas are merged. Just the intro and last scene are regular real-time footage. One frame required about 10 minutes of processing time (raw conversion + stacking). Unfortunately, the success rate was very low due to copious technical challenges and I spent almost 9 long months just to learn how to make these kinds of videos and understand how to work with these delicate creatures.

I am glad that I abandoned the idea of making this clip in 3D (with two cameras) – very few people have 3D screens and it doubles processing time.

Gear:
– Cameras: Canon 7D (died at the beginning of the project as I had overused it in my research), Canon 5d Mkiii (90% of footage is done with it)
– Lenses: Canon MP-E 65 mm lens, and a custom photomacrography rig (custom lenses are better for this type of task)
– Lights: adjustable custom-spectrum lamps (3 different models) – they were needed to recreate natural underwater illumination.
– several motorized stages, including StackShot for focus stacking. StackShot, is sadly not 100% reliable at all and kept destroying my footage.
– multiple computers to process thousands of 22+ Mpx raw images and perform focus stacking (an old laptop died on that mission after 3 weeks of continuous processing).

Edited in Sony Vegas, Adobe Photoshop CS6, Zerene Stacker, and Helicon Focus.
Music: Atmostra III by Cedric Baravaglio, Jonathan Ochmann and Zdravko Djordjevic.

=== Sharing/Use ===

Inquiries/licensing/press: find my contact details here: microworldsphotography.com/About

Please do not share this clip to promote or endorse marine aquarium industry. I simply want people to admire life, but not to be told to buy stuff, especially poses captive animals
More about using my videos:
microworldsphotography.com/Image-Use/Video-Use-and-Licensing

(consideration to buy a print from my website or to use the tip jar below the video is always welcome, but this option is better: secure.marineconservation.org.au/donate.php?campid=701900000006kqX)

first tests playing with #deepdream #inceptionism

A visualisation of what’s happening inside the mind of an artificial neural network.

In non-technical speak:

An artificial neural network can be thought of as analogous to a brain (immensely, immensely, immensely simplified. nothing like a brain really). It consists of layers of neurons and connections between neurons. Information is stored in this network as ‘weights’ (strengths) of connections between neurons. Low layers (i.e. closer to the input, e.g. ‘eyes’) store (and recognise) low level abstract features (corners, edges, orientations etc.) and higher layers store (and recognise) higher level features. This is analogous to how information is stored in the mammalian cerebral cortex (e.g. our brain).

Here a neural network has been ‘trained’ on thousands of images – i.e. the images have been fed into the network, and the network has ‘learnt’ about them (establishes weights / strengths for each neuron). (NB. This is a specific database of images fed into the network known as ImageNet image-net.org/explore )

Then when the network is fed a new unknown image (e.g. me), it tries to make sense of (i.e. recognise) this new image in context of what it already knows, i.e. what it’s already been trained on.

This can be thought of as asking the network “Based on what you’ve seen / what you know, what do you think this is?”, and is analogous to you recognising objects in clouds or ink / rorschach tests etc.

The effect is further exaggerated by encouraging the algorithm to generate an image of what it ‘thinks’ it is seeing, and feeding that image back into the input. Then it’s asked to reevaluate, creating a positive feedback loop, reinforcing the biased misinterpretation.

This is like asking you to draw what you think you see in the clouds, and then asking you to look at your drawing and draw what you think you are seeing in your drawing etc,

That last sentence was actually not fully accurate. It would be accurate, if instead of asking you to draw what you think you saw in the clouds, we scanned your brain, looked at a particular group of neurons, reconstructed an image based on the firing patterns of those neurons, based on the in-between representational states in your brain, and gave *that* image to you to look at. Then you would try to make sense of (i.e. recognise) *that* image, and the whole process will be repeated.

We aren’t actually asking the system what it thinks the image is, we’re extracting the image from somewhere inside the network. From any one of the layers. Since different layers store different levels of abstraction and detail, picking different layers to generate the ‘internal picture’ hi-lights different features.

All based on the google research by Alexander Mordvintsev, Software Engineer, Christopher Olah, Software Engineering Intern and Mike Tyka, Software Engineer
googleresearch.blogspot.ch/2015/06/inceptionism-going-deeper-into-neural.html
github.com/google/deepdream