Photogrammetry test - Doll 2

 As part of the continuation of testing to develop the pipeline for 705 I did another photogrammetry test using a different monster high doll. I thought that this test would go better than the last one due to the original doll having glittery skin. But this test actually turned out worse!

In the previous test I had something that was at least recognisable, but this time the software could not recognise anything of the doll. Here you can see the point cloud for the images I took of the doll. While I could tell at this stage it would need some clean up I didn't realise how bad it was. 

I generated a high-density point cloud and it's here that I could see I had massive problems. The doll has only picked up in small areas and the software can't match any points together. It has picked up the wall and the table really well, but not the actual subject, it has also staggered the doll so it almost duplicated in different slices. 

You can see here in the model solid just how bad it is, as it doesn't resemble a doll in the slightest. 

I'm not quite sure why this happened, I spent a long time trying to fix this and get any data out. I took a lot of images so I tried to use different combinations, as opposed to using all of them, while I did get different results from this, none were even close to being usable. You can see below the number of images I have taken. 

I think the issues are from taking the images and I need to find a better way to collect them for the doll tests. I may need to get closer to the doll to take the images so that it fills most of the screen. Or the other way around is to take a blank image of the background without the doll, I can then use this in software to tell it where to recognise. I am not sure how to do this yet (but I know it can be done) I have been recommended to watch a great video that is of Alex Alvarez where he talks through how to do this step by step. I need to watch this video before I do any more tests using the doll/puppets. 

While this test hasn't worked I have learnt more about photogrammetry and what not to do/ things I need to improve before I try again with capturing a doll. I will watch the video and try again, but for the next stages of testing and building the pipeline I am going to focus on getting an environment into VR. This is because of the time restraints to complete the MA course and because the doll/ puppet is proving considerably difficult to do, I will be better using my time to learn and achieve a result using the environment. I can come back to trying to capture the puppets at a later date.