Azure Custom Vision Service Reviews

Azure Custom Vision is the easiest way to classify images, it is the best option for artificial intelligence services, one of the reasons I choose this software is its wonderful user interface for training that happens to be compatible with any programming language, In addition, I can customize my own vision models which makes it even more attractive, the fact of being able to upload images and let the same application do that difficult job of identifying the objective of the photo.

I love that Microsoft creates this type of software that gives immediate solutions to existing problems, and with everything they invest in cybersecurity I feel free to use their services. I never find anything negative in this wonderful software, its updates are successful.

Do not stop using Azure Custom Vision is the best artificial intelligence tool.

I love being able to generate that experience in my clients since I can upload information on the Web safely and know that Azure Custom Vision will do an excellent job, the most important thing is to feel sure that information will not be leaked because it has data privacy, it is awesome as just uploading an image and tagging the precision is perfect. The ease in which I can handle multiple documentation API, API keys generate a lot of security. The other reason this software is my favorite is that I can access it from wherever I want.

For me, it's incredible that you got a fully-fledged custom vision service, and for free! It's sonmewhat magic that, getting a dataset on your own, uploading it to the as-a-service form of Custom Vision, and with a few more clicks, you have a trained model, and the training is done in seconds! Seems unbelievable for me. Not so long, you needed to do it by hand, and not as a service you can freely consume. It's priced accordingly.

This is (as of January 2019) a relatively new service that Microsoft offers inside the Azure Services umbrella, and it has a couple rough edges. For example, at the moment I can't export directly my model to Tensorflow Lite (when using Object recognition, I can't select the "compact" models). But I wrote them, and they told me that this very feature will be available in the near future, so this won't be an issue soon.

Can be used initially as-a service, but later ported to your Azure Subscription to get better quotas.

The business problem I am solving is to get an object recognition model to be used with our computer vision based apps.It solves it better than other alternatives (e.g. try to train the model yourself, doing the entire project of training in local, with a TensorFlow local installation, and with all the commands and scripts) the ease to use, out-of-the-box service ready to throw to it some images, and getting in matter of seconds a trained model.

We can train the model just by clicking and we do not require any knowledge of machine learning Algorithms

The model become overfitted easily this is the main problem with it

I will recommend others to use Azure Custom Vision Service once

Problem of training of large dataset can easily be done just by single click

The custom vision service is by far the easiest way to create classification or object detection models for images. The

The only way to really tweak the model the service gives you is to train more images. While this isn't a bad way, there are some other optimizations that can be done to get better performance.

The Custom Vision service is at least worth considering if planning to show how viable using image AI can be for a product or service. It performs very well and can even be used for production.

Image classification problems to automate workflows

Powerful features, mature API documentation, multiple usages,

UI console for training and testing,

API for training and testing in code, supporting multiple programming languages.

Good UI for showing the traninig, testing and model performances results.

Too many different API keys, including project key, API key, iteration key,

prediction key and training key. Could be improved to a more concise API

calling technique.

Difficult to tune the model for your customized usage or purposes.

Could be better improved in this situation.

Easy and convenient to use. But there isn't much space for tuning your custom model. You can change the model by feeding different training samples and different tagging. The model itself is powerful enough to adapt your customized changes or purposes.

Solving a computer vision and image recognition project using Azure Custom Vision Service.

Benefits: Finished the project and accomplished the project goal, save much time for tuning the model or building a deep learning model on my own.

Part of the datapipe line, feeding images into the custom vision service to do object recognition and tag classification and feed the result back into the database.

That you can customize and personalize your designs.

You can have trouble uploading the images at times. But this could also be due to size.

Keep focusing on your current process. and you will get closer to success for both the client and your company.

I can get my work done quicker than before.

I like how Azure custom vision service quickly recognizes differences between the images. I've played around with it, adding or deleting some things, some big things and some small things and it is able to recognize the differences. That is a pretty cool feature. On a more personal note, I have yet to test it out on those kids spot the difference games. That will be my next experiment. It would be pretty cool to use it on that and quickly spot the difference. I feel like it is fairly easy to use as well. It is easy on the eyes and the user interface doesnt look too complicated. I'm still a novice but I haven't seen anything that looked like it would be an issue for when I really need to use it but so far so good. Whenever I use a new software, softwares that usually dont pass the test can be seen right off the bat. The way it looks, the initial feel of how things work, the navigation of the software, how easy it is to navigate, how many clicks does it take to get you somewhere. All these things matter because it gives the user the impression right off the bat if it is clunky or a smooth software to use.

I'm not sure if there is anything I dislike about Azure custom vision service. I'm sure I will eventually see something I dont like but for now, I haven't experienced anything too in depth yet to find something I do not like. Usually software that doesnt pass the mustard are ones that doesnt feel good using it at the beginning. the way it looks when you turn it on, how cluttered it looks, how many buttons there are, how substantive and connected each click of the wy as you're working on it, all these things that a user does in the first 5 minutes gives us a lasting impression of the software.

I've only tried it on small things, not tried it on any projects yet as I have not had the need to. Right now, my experience with it has only been testing it out. As I commented on the earlier question, my next experiment is to try it out on those kids spot the difference games. That would be pretty cool to find all the differences. It'll make you look super stellar to the kids.

We have been using the Azure Custom Vision Service for close to a year now. We have leveraged it for AI to anaylze and tag images with metadata that can be brought back into OpenText Media Management utilizing a service we have written called Rich Media Analysis (RMA). We have been having good analysis results utilizing Custom Vision this way.

Continued improvement in documentation on how to build/leverage custom training using Azure Custom Vision Service.

For those that are looking for custom training in AI service providers, Azure Custom Vision Service is a very solid choice.

Automated metadata tagging of assets in our DAM.

When bulk-tagging of metadata in assets in a DAM, being able to do this in an automated fashion using AI can be a significant time savings for DAM librarians and archivists.

I like that the algorithms test themselves on the images I submit.

Not able to be used for detecting minor differences.

Easy to incorporate into projects and fast detection after it has been trained.

Using a Raspberry Pi to train for detecting objects. Works quickly after the training to identify items.

I liked that it easily customized my own state-of-the-art computer vision models for my unique use cases. I just uploaded a few labeled images and let Custom Vision Service do the work!

It was a little hard to understand when I first started using the product!

It allowed me to use my labeled images to teach Custom Vision Service the concepts I wanted it to learn!