Week 4 Assignment: 3D Scan My Shoe

In this assignment, I tried the photogrammetry technique to 3D scan my shoe. When I first heard about this method, it seemed quite easy and magical how just taking photos could make 3D models. However, when I really tried it, I realized that it was not that easy and could take forever. To successfully make a 3D scan using this method, it is important to choose the right object, take the right photos and be extremely patient. If you are interested about this topic, please read the followings.

Steps:
1. Find the Right Object

2. Take 100+ Photos

3. Create the 3D Model

4. Modify the 3D Model

5. Final Work & Reflection

A shoe (or something else for 3D scanning)

A DSL camera

Meshroom

Meshmixer

In photogrammetry, it is very important to choose the right object. To be specific, the object should not be too big or too small, fluffy, reflective, or transparent.

After knowing the rule, I chose the bottle of my hair care oil as its color was a frosted light brown which I thought fit into the rules. Moreover, I thought its shape was quite simple and the computer could quickly process its model out. However, I failed. After talking with professor Garcia, I realized that the shape of the object I chose, which was a cylinder, was the kind of shape that was easy for humans to draw using 3D modeling softwares but difficult for computers to recognize in photogrammetry. To further explain, this was because that every side of the cylinder looked so much the same, and the computer felt confused and could not generate the model.

Therefore, I decided to take photos of one of my shoe instead as it was in a more irregular shape with texture which was easier for the computer to recognize.

Taking photos is another very important step in photogrammetry. The key points for success including taking photos in an outdoor and cloudy area using a DSL camera turning to AV mode with 8.0 aperture and +-800 ISO. Moreover, it is better to have a piece of paper, with its color forming a contrast with your object, under your object.

I have took two sets of picture indoor using my phone and they all failed. After having the right setup which I mentioned above, I walked around the object and changed position little by little. I spent about 20 minutes and took 115 photos. That was, about 4 to 5 circles with different height around the object, and also some detailed photos for example at the logo of the shoe. After taking all these photos, we could start the next step.

In this phase, the only thing you need to do is upload all your photos into Meshroom and hit "start." The computer will process everything by itself. The process could take about 2-4 hours depending on the numbers of pictures you have.

Make sure that your computer have enough space which is more than 5GB or else the process will not complete. When I was in this phase, the computer I was using didn't have enough space which caused the program to stop until the professor cleaned up some unused software which provided the program with more spaces.

During the process, you could check on the structure form motion node. If it works, the possibility of you getting the right model you want will increase. If it doesn't work, you probably need to try taking another set of photos or change your object.

Later, you could double click on the meshing phase to see if the model is built in a right way. Many of my classmates had a good structure form motion picture but still failed on the meshing phase and got a weird model. I didn't know why this happened and I guessed it was possibly because their object had a similar color to its background and the computer found it difficult to recognize which part was the subject.

If the meshing phase build the right model, you are probably going to succeed. You just need to wait patiently to the last step and right click to download the obj file of your model.

After generating the 3D model by using Meshroom, we need to fix some errors and modify the model. I imported the shoe model into Meshmixer and used the plane cut function to cut the plane under my shoe. After that, I used the inspector function in analysis tools to find all the minor errors and fix them automatically. Finally, I used the smooth tool to smoothen some inside parts of the shoes.

I was satisfied with the look of the model and tried to export it as an stl file. Since I wanted to upload this file to instructables which only accept file less than 25MB and my file was larger than 25MB, I selected my object and use the reduce function to decrease the points it had and therefore fit into the range. Below, you could find that file.

The above are some pictures of the 3D model of my shoe. I was quite surprised how the final look was so delicate after experienced many times of failure on trying photogrammetry. Almost all the details on the shoe could be seen clearly and I was so happy about that.

In this assignment, I realized that 3D scanning could be a good choice when it comes to build some irregular shapes object. However, when it comes to easier shapes, it is better to do the modeling by oneself and don't choose 3D scanning because of laziness. The process isn't as easy as I thought before I tried it and really need our patience. Still, I view 3D scanning as a very amazing and super useful technique. I'm so glad of learning it and believe that it will help me a lot on my future projects!