The TED-Talk given by Blaise Aguera y Arcas impresses me a lot, it is unbelievable to me that in 2007 there is already an application that can connect 2D images together to reconstruct a 3D model. The way how it works is based on images from all over the world made in all ways that have different perspectives of a certain thing, and combining the details of this from all perspectives to reconstruct it in a 3D ourcome. But the problem is obvious, the quality of the images directly influence the 3D model, the authenticity, the sharpness, the depth of the image, there are many factors that can influence the reconstruction. As is shown in the presentation, there is a image that the Notre Dame de Paris is actually in a poster which is not real, but is also being counted in the application to form perspectives on the model, this might influence the accuracy of the reconstruction. Also we need to think of that there might be images that have been processed, these will make great influence that can cause distortion. I think this is the reason why it takes such long time, to deal with the classification of the images collected, which is related to machine learning, I think.
Talking about the evolvement of the challenge, what I have observed from the publications of the four original developers of the Photosynth, they are still in some way working in 3D image area. Many challenges are related to machine learning that is to better the identification in images, bettering the resolution the same as improving the ways of motion and image capture technologies. Those who still stick it out try to improve image and motion capture, the others work on making the environment of the 3D model more real, giving stereoscope to the environment, setting geo-localization. I think it is going to be impressive if this techonolgy becomes mature, as we can see 3D models built up based on related 2D images, for example, even though the Notre Dame de Paris has been burnt, we can still reconstruct it in 3D modeling based on past images. What’s more, we can use 2D imaginary pictures to built up 3D models for things that are in fantasies or myths or reconstruct the relics that has already been destroyed.