Comments (5)
Hi Julio!
It's been a while but if I remember correctly, gel_min_depth and max_depth are not the actual deformation of the gel. They are the distances from camera (bottom of the sensor) to deformed and undeformed gel. You can first try to measure these distances.
Hope it's helpful.
from digit-depth.
Hi, thanks for your quick answer.
I've managed to train the MLP and test it with a few images from my DIGIT sensor. However, the depth estimation is not correct when there is no contact applied to the sensor. Maybe you have experienced this issue and you can help or guide me to improve the quality of the depth estimation.
I attach the training images, the cvs file, the digit.yaml, and a few examples of the depth estimation. In the examples you will see that the model estimates an initial depth but there is no contact applied in that area.
from digit-depth.
That seems strange to me. It's been a while since I calibrated the sensors myself and I no longer have access to one, so I cannot test it on my own.
I would suggest doing another annotation and playing around with the gel parameters in the config file. Also, I noticed you annotated more around 700 images. Usually, for me, abour 30-40 images were enough to get good depth estimation. So, maybe, please try to use only 40 images in training the MLP model.
from digit-depth.
Hi, I am using the label_data.py script to label the images. I was annotating several points of the same circunference, that is why there were lots of rows inside of the csv file. Now, I am labelling only the center and one point of the circunference. Is that the right way to do it? I am getting the same depth estimation. Which gel parameters can be changed?
Thanks in advance.
from digit-depth.
Hi again,
Now I am getting better results. I think the problem is that the images need to contain contact regions in different positions of the gel. I will try to record more images with more variety of contact location and size to see if I can improve the results.
Now, I am trying to understand the mathematical equations to obtain surface normals from pixel values from these lines:
dist = np.sqrt((x - center_x) ** 2 + (y - center_y) ** 2)
ang_xz = math.acos(dist / radius)
ang_xy = math.atan2(y - center_y, x - center_x)
nx = math.cos(ang_xz) * math.cos(ang_xy)
ny = math.cos(ang_xz) * math.sin(ang_xy)
nz = math.sin(ang_xz)
but I am struggling a lot. I have also read the gelsight article but I can't see how to get these values.
from digit-depth.
Related Issues (9)
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from digit-depth.