The previous post examined the processing of photographs of a cultural artefact to make a 3D model using Agisoft Photoscan Professional as part of the museum case study in the NUIM-AFF621: Remaking the Physical – 3D recording assignment. This post will look at 123D Catch for making a 3D model from the same captured data. The captured images are of a bipartite bowl (2004E505:5) found in a cist grave in Liscooly, Co. Donegal in 2004. The images are courtesy of the National Museum of Ireland (Archaeology) who supported the project and provided access to the artefact. A detailed description of the artefact is presented in Post 2.
From the Autodesk website, I easily found the download for 123D Catch and it was straightforward to install. I briefly looked at the 123D Catch quick start guide, and the accompanying videos, it seemed simple to use, so I was quite happy to proceed. Beginning with “Create New Capture”, the user is prompted to log in or sign up as a new user. The reason for this is that the user will need to log into an Autodesk account to access the final 3D model on the Autodesk website. Moreover, the user needs to log in to make any adjustments to the project model prior to submitting the scene to create a final 3D model online. The next step is to upload the photographs; however, there is only a maximum of 70 allowed. As I had 138 photographs in a master file from the capturing of the Liscooly bipartite bowl (2004E505:5), I created a separate folder for the photographs I was going to use in 123D Catch. Without really thinking it through, I randomly selected 70 photographs from the master folder and copied them to the new 123D Catch folder and uploaded again. Once uploaded, I clicked on “Create Project” and was prompted to fill out some details – such as my email address, and the title and a description of the model. Then it was a simple matter of waiting for 10-15 minutes.
Initially, I was surprised at how easy the process seemed, especially in comparison to Agisoft PhotoScan. Although, it was similar in terms of demands on the system as I received regular notices indicating high disk/memory usage by the programme. I was somewhat excited waiting for the final 3D model to load on the website, half-expecting a 3D masterpiece – though that was short lived as my first 3D model had a double surface on one side, a large hole on the other side, and no outer base – obviously something was wrong. Having watched some 123D Catch tutorial videos on YouTube, some instructors indicated that 40-50 photographs was sufficient to create a 3D model. So, I thought that perhaps I had too many photographs and I deleted some (randomly again) from the 123D Catch file. I ended up with about 50 images and I went through the process again of uploading the files and creating a new project. There was less waiting time involved this time around but the 3D model produced was equally distorted. So, it was back to the Internet to watch some more tutorials.
From various online sources, I came to understand that the 123D Catch software algorithm tends to prefer captured images of similar distance, and evenly spaced. In the image above, it is obvious that the photographs I uploaded were not evenly spaced, and some concentrated too much on one area. Furthermore, some of the photographs were not included in the final 3D model as apparently they did not auto-stitch, but it took me a while to figure that out (more about this further on). Also, I had not clicked on 2-sided model in the settings; thus, there was no base in the 3D model.
I went back to the drawing board and set up a new 123D Catch folder for the photographs I was going to use in the next attempt. I selected 62 photographs from the master folder that reflected an even spacing, and a similar distance. I uploaded for the third time and my very first 123D Catch model which actually looks something like the Liscooly bipartite bowl is below. However, while it was an improvement on the first two attempts, the 3D model is extremely dull in comparison to the original vessel, and even though I clicked for a two sided-model, the base is still distorted.
3D Model Title: Version3_3D_bipartite_bowl_2004E505_NMI
(click here to view in 123D viewer)
Still working with the same 62 photographs, I decided to alter their brightness and contrast through the Microsoft auto-correct image tool; I uploaded to 123D Catch again, and waited. However, the resulting 3D model (below) appears to be over-textured in places, the base is still incorrect, and overall, it is not a very good representation of the original vessel.
3D Model Title: Version4_3D_bipartite_bowl_2004E505_NMI-autocorrect
(click here to view in 123D viewer)
For the fifth attempt, I decided to start again.
I selected 68 photographs from the original master file which were of equal spacing and of similar distance. Then, I rectified the brightness and contrast of the images manually with the Microsoft image tool, rather than apply auto-correct. This took some time, but I thought that I should at least try to even out the texture for all the photographs. Then, I created a new project and crossed my fingers hoping that this would be the final attempt.
It was only in this attempt that I realised that some of the images did not auto-stitch in the project model, this is indicative by a yellow triangle with an exclamation mark. I then gathered that this had probably happened in all previous attempts! Indeed, the photographs that did not stitch were mostly for the base of the bowl. So, I found a tutorial on YouTube for manual stitching in 123D Catch.
Determined to produce a 3D model that would justify a representation of the Liscooly bipartite bowl (2004E505:5) – I began the process of manual stitching the photographs that did not auto-stitch. This was no easy task, and very time consuming due to the repetition of the design of the vessel. However, during this process, I was prompted to sign in again, in order to proceed as my broadband connection indicated that it was limited. Moreover, I could not continue the manual stitching process until my Internet connection improved. The next morning, with the Internet back at full strength, I opened the same capture scene and decided to submit it for processing as a tester to see if there was any improvement with the few photographs that I did manage to manually stitch successfully. Of the 62 photographs uploaded for this scene, 45 auto-stitched photographs and a handful of the manually stitched photographs were submitted for a test-processing.
The model above shows a distinct improvement, though the base is still distorted, however, at this point I believed this would be resolved by manually stitching the rest of the photographs of the base that did not auto-stitch in the first instance. So, I proceeded and finally completed the manual stitching and submitted the scene for processing.
More dilemmas – just as it seemed to upload, I received a message on screen as follows.
I joined Autodesk 123 Support Forum to see if anyone else had similar issues, and apparently other people have experienced the same error, and in particular from submitting the scene from manual stitching (one report was just 12 days ago – error: unable to access the remote server). I left a message and received one response from another user on 02 April who was experiencing the same.
I was disappointed, but decided to give it one last try. I double checked there was enough overlap of the images, I double checked the brightness and contrast, and I created a new project with 62 photographs and waited for the application to create the scene. I then manually stitched the photographs that did not auto-stitch, and these were mostly of the base of the vessel. I double checked the manual points once more and I submitted the scene for processing, but it was not to be as I got the same error message again. This is unfortunate as other than this I was impressed with 123D Catch in terms of its ability to produce a 3D model with less effort for the user in comparison to Agisoft PhotoScan. Additionally, 123D Catch allows for the download of an entire project, that is inclusive of the programme file, the photos, the mesh and other other related files (obj. mtl.). Hence in terms of the preservation of data and methodological transparency, other users can access the files online for further research and to remix datasets.
While I started this endeavour with high hopes and initially perceived it to be a simple process, this was not the case. It was time consuming and frustrating. However, in fairness, much of the problems I experienced with 123D Catch originated with the captured data, and so, that is certainly my fault. Had the images been more consistent in the first place, in terms of brightness and contrast, the application may have been able to auto-stitch them without the necessity for manual stitching. Moreover, had I uploaded images that were evenly spaced and of a similar distance in the first instance, I would not have wasted so much time in trying to figure it out. From this experience I will conclude with some main points to remember when using 123D Catch:
- Check that the computer being used has enough memory capacity to deal with the application;
- Ensure that there is a strong Internet connection;
- Input captured images that are evenly spaced, of a similar distance, and of a consistent texture; thus, ensure you capture the best photographs possible;
- Carefully – manually stitch photographs that do not auto-stitch in the opening scene process, and double check the manual points before submitting the scene for processing; however, unfortunately, there is presently an error with the application when submitting a scene for processing from the ‘manual stitch’ interface.
At this point, I needed to submit a record of the digital files I created for 123D Catch, and still no resolution for the processing of a scene from manual stitching, therefore, I am submitting the last files which can be seen online as part of my digital file submission for this part of the assignment. The project files used in this case study can also be downloaded from here.
Update 12/04/2015: I got a response from a 123D Catch Official Rep concerning my query with the ” unable to access the remote server” problem in manual stitching. Apparently, there is still an issue with manual stitching – Response is available here.
- Noursalehi, Ehsan. “123D Catch – How to Make 3D Models from Pictures.” YouTube, 07 May, 2012. Web 24 Mar. 2015. <https://www.youtube.com/watch?v=NsBg-m2hrIM>
- 123D Catch. “123D Catch – Create Your Own 3D model: Planning your shoot.” YouTube, 05, November, 2011. Web. 24 Mar. 2015. <https://www.youtube.com/watch?v=7TfXXJxDsXw>
- 123D Catch. “123D Catch – Navigating your model.” YouTube, 04, November, 2011. Web. 24 Mar. 2015. <https://www.youtube.com/watch?v=4GvbO3z0mpE>
- 123D Catch. “Mesh Details.” YouTube, 11, November, 2011. Web. 24 Mar. 2015. <https://www.youtube.com/watch?v=4GvbO3z0mpE>