Stereoscopic panoramas are sparse. Due to difficulties with lenses and stitching, the norm for the average user is to only have monoscopic -2D- panoramas. The alternative is an expensive camera and an elaborate stitching process.
Deep360 takes another approach.
Deep360 converts your monoscopic equirectangular panoramas to 3D with the use of Artificial Intelligence. It estimates the depth using scene topography and outputs a new image with the perspective offset for stereoscopic viewing.
Required specs currently: 64 bit Windows 7/10 with 4GB RAM minimum.
Required specs for video support will be: Core i5 or higher, 16GB RAM, GTX/RTX NVidia videocard (6GB of RAM, 4GB might suffice too). Processing time for 15 minutes of video at 24 fps will take from 2 to 4 hours. The video version will be a licensed version, the single shot version without Cuda support will remain free.
Deep360 is currently available for free. It is released as v1 but regard it as an alpha release.
I would love to receive your input, requests, bug reports or any other feedback by email or through other channels.
Deep360 does not log any information, contact any servers or make any changes to your system. However note that I'm not accountable for any problems that may occur by running this software.
There's a bug in v1.2Beta which appears with batch processing.
v1.3: has another way of processing the image by default. Switch to the previous way with the 'Alternative Mode' checkbox. Clearing of cache is needed.
v1.4: Renamed Alternative Mode to Accurate Mode and increased accuracy
v1.5: Improved memory consumption and speed for non-Accurate Mode. Tested with 16Kx8K image.
v1.6: Improved speed. Added setting to set convergence (0 is focus at infinity). Removed legacy mode. Set dilation to 0 for additional speed boost.
My name is Kaj Toet. HoloKilo is my brandname which is used for a multitude of 3D related products. You can view my portfolio at kajos.net and you can contact me at the email address shown in the image below.