I have been working on some 'experimental testing' of the applications to create base models. The models are below. I will also detail a new way to test the models and applications so that we get the best results. The goal is to standarize the test so that we test same conditions for all methods. This will allow us to make a good comparison between all of them for us to pick a model.
- LumaAI:
- Photogrammetry Model: based on a video or footage that you already have. I did this in the web. The video I used is in the video folder at the lab's computer on the desktop. Here is a link that model:https://lumalabs.ai/embed/3c6fbcbe-c0d1-4add-88dc-17a8ba73cdb8?mode=sparkles&background=%23ffffff&color=%23000000&showTitle=true&loadBg=true&logoPosition=bottom-left&infoPosition=bottom-right&cinematicVideo=undefined&showMenu=false
- NeRF: used my phone and their guidance lines to create a model. This is done with a phone. Here is the model to that: https://lumalabs.ai/embed/096fc605-82e2-4068-992a-0583c08642b9?mode=sparkles&background=%23ffffff&color=%23000000&showTitle=true&loadBg=true&logoPosition=bottom-left&infoPosition=bottom-right&cinematicVideo=undefined&showMenu=false
- Polycam:
- LiDAR: Here is the model with the phone application and the LiDAR scanner. https://poly.cam/capture/52ED5AD5-F7EB-46FD-9CB3-7AFBBBACD673
- Photogrammmetry: The model is similar to one that Reality Capture would yield with this input. The input was a 4 minute video. Here is the model: https://poly.cam/capture/ec68f83e-a2b6-44c7-847d-d13249dad2fa
- Gaussian Splatting: This model failed, but the application did not tell me a reason why. Plan is to record a video to hand it over to Professor Clark's pro version of the app. Here is the link: https://poly.cam/capture/68dcbbfc-e161-4862-aa6e-812d64b06615
- Testing moving forward:
- I will create a video of the office that has the following:
- Same lighting conditions(I will bring in an extra light and maybe a light diffuser so that I can reach the dark areas too)
- Same length video(will create a 10 minute video with detail)
- Same position of the objects within the office
- No people inside the office
- If the video is not an input:
- Same positiono of the objects within the office
- No people inside the office
- Same lighting conditions
- I will create a video of the office that has the following:
- Next Steps:
- Record the video described above
- Create models with the descriptions
- Create a photography set of images in the lab of 200 images max. Then use this in Reality Capture and other methods.
- Give video to Professor Clark.