Google Cloud AutoML Vision Pros and Cons

Algorithms are so much better than other ML platforms. Really great for Machine learning and so easy to train and so fast to use. Review collected by and hosted on G2.com.

Everything is headed in the right direction. Keep up the awesome work! Review collected by and hosted on G2.com.

I was helpful thing is the best learning techniques when learning to use machinery and one of the best things about this is that it is precise and accurate, and that I am confident I am learning correctly Review collected by and hosted on G2.com.

What I dislike the most, is the training models they could be more accurate and more detailed. This could be better by putting in more detail and more realistic images with in the training models. Review collected by and hosted on G2.com.

It supports serving models to edge devices with different hardwares. I had optimized an image classification model for ARM Mali GPU, It was good to have GCP support to various hardware platforms Review collected by and hosted on G2.com.

In realtime it is difficult to monitor the models at production. Need to rely on third party software which is a bit tedious. If there are any GCP services that I could use to integrate with automl, it would be benefecial Review collected by and hosted on G2.com.

Google Cloud AutoML vision is pretty to use and developer friendly. We can also save a lot of time on training models since ML vision is ready to be used out of the box. Review collected by and hosted on G2.com.

Cost. long term cost is pretty high once the volume accumulates. For huge volume, it may make us to start exploring alternative solutions or even building our own models Review collected by and hosted on G2.com.

A lot of precision compared to other OCR engines, in addition to having a very fast response time. Another advantage is that there is no need to process the image before uploading it, unlike the other platforms I had used, which saves a lot of work time. Review collected by and hosted on G2.com.

Perhaps what I liked the least was the fact that there is little documentation on the subject to be able to understand the errors well. In short, there is little information to be able to learn. Review collected by and hosted on G2.com.

Google autoML is an Open source tool for AI automation

It supports REST API

24/7 technical support with GCP team

Train custom machine learning models

Detect objects, where they are, and how many Review collected by and hosted on G2.com.

Premium pricing but can get discounts based on the requirement & usage

limitation in linking defferent algorithms

limited to video, image & text as of now

could not have option to custom integrate other 3rd party ML Review collected by and hosted on G2.com.

It offers a lot of services apart from just saving time by taking care of most of the Machine Learning Modelling on its own. One of the features I like the most is the ability to get the model offline as a docker image or Tflite rather than having a server endpoint. It helps to deploy the model on Edge devices and help it run offline.

During training, we get the option to select the model type to be more accurate or have better accuracy, so we can decide the best Time(Latency vs Accuracy ) tradeoff based on our personal needs.

It also notifies via email when training is complete and shows models metrics like precision, recall, and accuracy after training. Review collected by and hosted on G2.com.

It is quite easy to use the API because of the user-friendly GUI and not having to use packages to run the models. I have not faced any problems and everything has worked well for me till my use. Review collected by and hosted on G2.com.

What I like most is that it allows me to classify custom images and practice with them. It is good for implementing models of different latencies and optimizing them. Review collected by and hosted on G2.com.

What I like least about this program is that it depends on various other programs that have a high cost, I understand that using Google Cloud requires payment but it depends on third-party programs. Review collected by and hosted on G2.com.

I like that it is a simple and easy-to-understand tool to quickly carry out the construction of models for tasks without the need for coding. Review collected by and hosted on G2.com.

Complex data types, unsupervised learning, and reinforcement learning. Review collected by and hosted on G2.com.

It helped me in quickly creating models for certain classification task witout using a single line of code and additional to that was getting great suggestion from model improvement Review collected by and hosted on G2.com.

The most important thing that I dislike about was that I was getting poor prediction performance. And sometimes the images did not load properly. Apart from that it also seemed to have an overall slow response Review collected by and hosted on G2.com.