This basically is a cubemap in a skybox (cube with camera at the center), there are also spherical ones (in a sphere obviously) but those having bad quality at the poles and require way more polygons.
There is a THREE example:
For the images you require multiple images you'll take from one spot rotating the camera each shot 360°, but there are also apps to create the images with a smartphone, but a tripod is really required, one displaced shot and it gets blurred. Then stitch those together in a graphics software like photoshop and/or use out-of-box tools to generate the image out of it. For good performance and resolution most viewer using LOD techniques and split the source image into tiles for multiple resolution (mobile/desktop), to lazy load tiles inside the frustum and to load zoom steps.