In the previous reality capture system users could set a parameter which would determine the resolution of the output models. I want to wind up with models about 100-150K vertices. Is there a setting that allows me to request the modeler to keep the number of generated vertices within some bounds, somewhere in the forge API?
The vertex/triangle decimation is usually, what can be called "subjective" task, which can also explain why there are so many optimization algorithms in the wild.
One type of optimization you would need and expect for "organic" models, and totally different one for an architectural building.
The Reality Capture API provides you only with raw Hi-Res results, avoiding "opionated" optimizations. This should be considered just as a step in automation pipeline.
Another step, would be, upon receiving, to automatically optimize the resulted mesh based on set of settings you need.
One of these steps could be Design Automation for 3ds Max, where you feed a model and using the ProOptimizer Modifier within 3ds Max, you output the mesh with needed detail. A sample of this step, can be found here: https://forge-showroom.autodesk.io/post/prooptimizer.
There are also numerous opensource solutions which should help you cover this post-processing step.
Related
I have been trying to tackle a problem where I need to track multiple people through multiple camera viewpoints on a real-time basis.
I found a solution DeepCC (https://github.com/daiwc/DeepCC) on DukeMTMC dataset but unfortunately, this solution has been taken down because of data confidentiality issues. They were using Fast R-CNN for object detection, triplet loss for Re-identification and DeepSort for real-time multiple object tracking.
Questions:
1. Can someone share some other resources regarding the same problem?
2. Is there a way to download and still use the DukeMTMC database for multiple tracking problem?
3. Is anyone aware when the official website (http://vision.cs.duke.edu/DukeMTMC/) will be available again?
Please feel free to provide different variations of the question :)
Intel OpenVINO framewors has all part of this task:
Objects detection with pretrained Faster RCNN, SSD or YOLO.
Reidentification models.
And complete demo application.
And you can use another models. Or if you want to use detection on GPU then take opencv_dnn_cuda for detection and OpenVINO for reidentification.
A good deep learning library that I have used in the past for my work is called Mask R-CNN, or Mask Regions-Convolutional Neural-Network. Although I have only used this algorithm on images and not on videos, the same principles apply, and it's very easy to make the transition to detection objects in a video. The algorithm uses Tensorflow and Keras, where you can split your input data, i.e images of people, into two sets, training, and validation.
For training, use a third party software like via, to annotate the people in the images. After the annotations have been drawn, you will export a JSON file with all annotations drawn, which will be used for the training process. Do the same thing for the validation phase, BUT make sure the images in the validation have not been seen before by the algorithm.
Once you have annotated both groups and generated JSON files, you then can start training the algorithm. Mask R-CNN makes it very easy to train, with all you need to do is pass one line full of commands to start it. If you want to train data on your GPU instead of your CPU, then install Nvidia's CUDA, which works very well with supported GPUs, and requires no coding after the installation.
During the training stage, you will be generating weights files, which are stored in the .h5 format. Depending on the number of epochs you choose, there will be a weights file generated per epoch. Once the training has finished, you then will just have to reference that weights file anytime you want to detect relevant objects, i.e. in your video feed.
Some important info:
Mask R-CNN is somewhat of an older algorithm, but it still works flawlessly today. Although some people have updated the algorithm to Tenserflow 2.0+, to get the best use out of it, use the following.
Tensorflow-gpu 1.13.2+
Keras 2.0.0+
CUDA 9.0 to 10.0
Honestly, the hardest part for me in the past was not using the algorithm, but finding the right versions of Tensorflow, Keras, and CUDA, that all play well with each other, and don't error out. Although the above-mentioned versions will work, try and see if you can upgrade or downgrade certain libraries to see if you can get better results.
Article about Mask R-CNN with video, I find it to be very useful and resourceful.
https://www.pyimagesearch.com/2018/11/19/mask-r-cnn-with-opencv/
The GitHub repo can be found below.
https://github.com/matterport/Mask_RCNN
EDIT
You can use this method across multiple cameras, just set up multiple video captures within a computer vision library like OpenCV. I assume this would be done with Python, which both Mask R-CNN and OpenCV are primarly based in.
I am working on a Wysiwyg Editor for CesiumJS content.
The user will be able to create many points, lines and other graphics, connect them according to definable relations and group them in separate Groups.
Now I am wondering what the best practises are in terms of performance.
At the moment I create one PointPrimitiveCollection for each Group
and then add points:
group.points = scene.primitives.add(new Cesium.PointPrimitiveCollection());
and then
group.points.add({
position : cartesian,
...
});
for each new point.
Polygons are created using:
network.hull_polygon = viewer.entities.add({
name : 'xxx',
polygon : {
hierarchy : Cesium.Cartesian3.fromDegreesArray(points_array),
material : color,
...
}
});
polylines similarly.
Now since the Objects can also be dragged around / animated, I was wondering where Cesiums entity logic would come in?
Thanks for all help!
Cesium's Entity logic is useful primarily for objects that move along a known path over time, for example the flight plan of an aircraft in the future, or a GPS recording of the route taken by a vehicle in the past. Such routes can be loaded into the Entity system (often via CZML), and the user can run the simulation time forwards and backwards at arbitrary speeds, to review the routes of all the vehicles. The Entity system owns the logic for updating graphics primitive positions based on simulation time changes.
Entities are also often used as a quick way to make some disparate graphics primitives associate with each other. For example, a polygon, a point, and a label can all be created as a single Entity even if they are three separate graphics primitives at the same location. This saves a bit of effort on the part of the application developer, and doesn't hurt performance too much since the properties involved are all marked as constants, so the Entity layer knows not to update them with simulation time.
But, it sounds like you may have a case where paths are not known in advance. For things like user interactive edits or real-time telemetry being received, the Entity system can't know what's coming up next, so its whole system for updating positions from simulation times is not doing you any good. In that case it may be better to skip the Entities, and deal exclusively with graphics primitives for this. This would mean you need to write your own update function to alter graphics positions as new information is being received, similar to the Entity layer's update functions, but based on your own live inputs instead of recorded paths.
Note that the public "Sandcastle" demos only include Entity demos. But, if you download and build the source for Cesium and run Sandcastle locally from a dev build, a separate tab appears in the Sandcastle Gallery called Development that shows a whole set of demos based on graphics primitives as opposed to Entities. This can be useful for seeing examples of how to control things at this layer.
Hopefully this is helpful in understanding how the different layers of Cesium interact.
I need to visually recognise some flat pictures showed to camera. There are not many of them (maybe 30) but discrimination may depend on details. The input may be partly obscured or shadowed and is suspect to lighting changes.
The samples need to be updatable.
There are many existing frameworks for object detection, with the most reliable ones depending on deep learning methods (mostly convolutional networks). However, the pretrained models are not well optimised to discern flat imagery of course, and even if I start training from scratch, updating the system for new samples would take a cumbersome training process, if I am right about how this works.
Is it possible to use deep learning while still keeping the sample pool flexible?
Is there any other well known reliable method to detect images from a small sample set?
One can use well trained networks for visual classification like Inception or SqueezeNet, slice of the last layer(s) and add a simple statistical algorithm (for example k-nearest neighbour) that can be directly teached by the samples in a non-iterative fashion.
Most classification-related calculations like lighting and orientation insensitivity are already handled by the pre-trained network then, while the network's output keep enough information to allow statistical algorithms decide the image class.
An implementation using k-nearest neighbour is shown here: https://teachablemachine.withgoogle.com/ , the source is hosted here: https://github.com/googlecreativelab/teachable-machine .
Use transfer learning; you’ll still need to build a training set, but you’ll get better results than starting with random weights. Try to find a model trained on images similar to yours. You might also do some black box testing of the selected model with your curated images to baseline it’s response curve to your images.
I'm using Autodesk Forge to integrate with our remodeling tool. In particular, I need to count objects of different families and types and determine to what room they actually belong. I use Model Derivative API for this purpose. To keep the room/area information I convert .rvt files to .nwc files as suggested here. However, when I retrieve data with GET /modelderivative/v2/designdata/{urn}/metadata/{guid}/properties I face the following problems from time to time:
Room information sometimes disappears from Objects for some reason
Objects disappear from result data for some reason (but they seem to exist when I browse them in A360)
I have no idea, what can be the reason for this.
I have no explanation for the disappearance of room data or objects for you.
If you can provide a reproducible case demonstrating that, I will gladly pass it on to the development team for analysis.
If you are interested in an immediate reliable solution and full control, which I assume is the case, I would suggest following the second bullet item in the advice provided by Eason in the previous answer that you refer to above:
Extract all the room information and object relationships you are interested in via the Revit API, store that data somewhere yourself, and use it later on wherever you like to your heart's content.
Then you will be completely safe and independent of all other components and their unpredictable behaviour.
If the only information that you need is the room containing each family instance, I can even implement a suitable Revit add-in for you.
Another suggestion that might help, if that is indeed the data you require: determine that information in a Revit add-in and attach it to each family instance in your own personal shared parameter. That will ensure that it remains intact through the translation process. Afaik, all shared parameter data is retained, independent of other behaviour.
Can machine learning be used to transform/modifiy a list of numbers.
I have many pairs of binary files read from vehicle ECUs, an original or stock file before the vehicle was tuned, and a modified file which has the engine parameters altered. The files are basically lists of little or big endian 16 bit numbers.
I was wondering if it is at all possible to feed these pairs of files into machine learning, and for it to take a new stock file and attempt to transform or tune that stock file.
I would appreciate it if somebody could tell me if this is something which is at all possible. All of the examples I've found appear to make decisions on data rather than do any sort of a transformation.
Also I'm hoping to use azure for this.
We would need more information about your specific problem to answer. But, supervised machine learning can take data with a lot of inputs (like your stock file, perhaps) and an output (say a tuned value), and learn the correlations between those inputs and output, and then be able to predict the output for new inputs. (In machine learning terminology, these inputs are called "features" and the output is called a "label".)
Now, within supervised machine learning, there is a category of algorithms called regression algorithms. Regression algorithms allow you to predict a number (sounds like what you want).
Now, the issue that I see, if I'm understanding your problem correctly, is that you have a whole list of values to tune. Two things:
Do those values depend on each other and influence each other? Do any other factors not included in your stock file affect how the numbers should be tuned? Those will need to be included as features in your model.
Regression algorithms predict a single value, so you would need to build a model for each of the values in your stock file that you want to tune.
For more information, you might want to check out Choosing an Azure Machine Learning Algorithm and How to choose algorithms for Microsoft Azure Machine Learning.
Again, I would need to know more about your data to make better suggestions, but I hope that helps.