Couldn't you also serialize the resulting model to JSON (including weights/values), stick that JSON hard-coded into your source, and deserialize at runtime to execute the model? This is how I typically deploy neural networks and other machine learning models.
For example, http://colorbot.herokuapp.com stores the optimal model for the neural network in JSON, right in the code.
It's not too hard to imagine a user group that is good at predictive modeling, but bad at setting up a reliable service (mathmaticians, statisticians, other similar disciplines). These users may be familiar with R or numpy, but unable to build a web API quickly.
If we are uploading cheques and licenses, why not just have a form with the respective upload sections for each document. Sure, people might get the two backward by accident, but is that error rate really any worse than the false positive rate of the proposed solution.
If you are going to go the machine learning route, then it really only makes sense if you go all the way and do some OCR on the documents to pick out the meaningful information to prepopulate some input screen that would likely appear after submitting those documents.
Yes, content analysis can be greatly simplified using symbolic inputs. GPS Tags in images, User input in OP's check example and most industrial pipelines used sophisticated barcode scanners to scan objects very quickly, but the cameras are at a fixed distance/height from the object and the objects are aligned in a particular direction. Some purists would argue that this is not computer vision, but it makes for some excellent engineering, and solves some hard problems very well.
I find Computer vision so interesting. Does anyone know if there are many consulting opportunities dealing with it?
Or is it a really crowded fields because it's so interesting to so many people. (I never see people looking for computer vision researchers in the who's hiring threads)
I used to work as a machine/computer vision engineer but that was before the web was big so the image retrieval opportunities didn't exist.
I worked mostly in small specialty companies doing inspection systems (sorting, metrology, controls, robotics) and then for many years at HP in manufacturing. In those positions computer vision was part of the job so it often didn't show up in the initial keyword searches. Most of the jobs required some additional skills, most often optics and some controls or automation.
If you broaden your search to include robotics and automation you may find more. Also, medical imaging is fascinating and of course there's defense but I didn't want that.
My company is always interested in talking to engineers with experience in computer vision and machine learning. We're in the advertising industry, doing large scale automated video content classification. Drop me a line if you're interested - email is in my profile.
11 comments
[ 0.22 ms ] story [ 38.7 ms ] threadyhat is built for embedding these types of analyses in a production application
For example, http://colorbot.herokuapp.com stores the optimal model for the neural network in JSON, right in the code.
If we are uploading cheques and licenses, why not just have a form with the respective upload sections for each document. Sure, people might get the two backward by accident, but is that error rate really any worse than the false positive rate of the proposed solution.
If you are going to go the machine learning route, then it really only makes sense if you go all the way and do some OCR on the documents to pick out the meaningful information to prepopulate some input screen that would likely appear after submitting those documents.
Or is it a really crowded fields because it's so interesting to so many people. (I never see people looking for computer vision researchers in the who's hiring threads)
I worked mostly in small specialty companies doing inspection systems (sorting, metrology, controls, robotics) and then for many years at HP in manufacturing. In those positions computer vision was part of the job so it often didn't show up in the initial keyword searches. Most of the jobs required some additional skills, most often optics and some controls or automation.
If you broaden your search to include robotics and automation you may find more. Also, medical imaging is fascinating and of course there's defense but I didn't want that.
Where I work: http://set.tv