Google Cloud AutoML Reviews

I really like the easy way to make ML Models. Even for me as a CEO

It is still in Beta and sometimes buggy.

we build an app when it was in alpha for a customer to detect their hinges

I like the google sheets/docs. it is a great way to collaborate.

google hangout is not as good as slack because you can't upload documents

it helps us create documents and collaborate.

Google's AutoML minimizes the hand-tuning of the hyper-parameters of machine learning algorithm, such as neural networks, deep networks, to near zero. Given a dataset of interest, AutoML takes the dataset of interest from the users and determines a high performing network architecture.

Given that the AutoML is proprietary software from Google, it is kind of Blackbox to the users, while understandable, users from the same domain kind of benefit from the nuances of the AutoML to deduce more intuitive inferences from insights used by AutoML to determine the network architecture.

AutoML has helped reduce the computational time and resources required to determine architectures for large datasets, such as ImageNet (ILSVR2015) which contains 1.25 million images of 220x320 resolution from real-world scenarios. With that being said, for a dataset of this magnitude, it is near impossible to hand tune effectively hyper-parameters.

Google AutoML kit is one of the best platforms for developers of Machine Learning, as the main benefit is that it is from Google and ML is highly affected by Google. In that manner, Google's AutoML kit provides very sophisticated ML techniques and algorithms support. In addition to that, the performance and accuracy of AutoML algorithms are way better than other ML platforms. Create and train the model of any specific problems is very easy and fast. Google AutoML Algorithms and techniques can easily handle large and complex data processing and as is it on the cloud, I get extremely good GPU support that would rather cost me thousands of bucks. It really makes the difference in how I develop ML applications.

One thing I believe can be improved that Google can get know regarding my private data sets which sometimes can be a privacy issue for some users. So, Google may introduce some private data loading techniques on AutoMl platform. Other than that there are not any major issues in it.

Google Could AutoML is a very cost-effective solution in terms of hardware resources. Also, with that, you will get security as well as reliable performance for any kind of complex data you want to process. Definitely anyone who is working in ML should go with it once.

I use Google Could AutoML for various purposes such as data preprocessing, dataset manipulation and mainly for model training. We have monthly subsciption based plan for Google AutoML platform at our company and as a machine learning developer, I am utilizing it for developing a different pilot application for demo purpose as well as main projects of ML in it.

Traditionally the major problem with ml apis have been that the services have been too generic for enterprise use. Automl enables us to finally use ml api services in an enterprise context by tailoring the results to the specific use case needs

Would be good to have ml automl recommendation engine

vision automl

Ease of use and ability to hire interns to tag photos so senior staff can focus on tech.

Hard for us to debug bad imagery matches.

In our case of image recognition, we needed 10s of thousands of pictures.

We are modernizing a laser tag system to instead use image recognition.

using cloud auto ml as a building block for complicated computer vision problems. How quickly it trains, clearly doing a great implementation of transfer learning.

that i cannot take my model out. That i am tied to making predictions through the api of this service. There is no clear upfront description of data ownership, is google reserving the right to look at my training data for future model building?

many problems related to earth sub-surface characterization.

The vision and text API's were really easy to use, or train on custom datasets. very good tutorials and help available online and supportive community

The late limits are little too small, even for students

Easy to set up, if at a hackathon, would recommend using this service.

I was trying to build a image recognition application at a Hackathon

That has a lot of services and I especially like the logging.

The documentation can be chaotic and daunting. I sometimes loose myself there.

The deployment of my models, the execution of them and also the data hosting comes from google. It helps to have all the tools I need in one place and they can interact with each other without issues and more configuration from my side.

Isso ajudou nossa empresa a melhorar a tomada de decisões com mais rapidez e facilidade. Diante de um conjunto de dados de interesse, o AutoML retira o conjunto de dados de interesse dos usuários e determina uma arquitetura de rede de alto desempenho. É a maneira rápida e fácil de começar a trabalhar rapidamente e obter modelos sofisticados mais tarde. Facilidade de uso e capacidade de contratar estagiários para marcar fotos para que a equipe sênior possa se concentrar na tecnologia

Ainda há muita arte associada aos modelos de aprendizado de máquina de treinamento Difícil para nós depurarmos correspondências de imagens ruins Everynow e então o beta tem uma falha que eu não posso tirar meu modelo

Eu estava tentando construir um aplicativo de reconhecimento de imagem em um Hackathon Estamos modernizando um sistema de marcação a laser para usar o reconhecimento de imagem, criamos um aplicativo quando estava em alfa para um cliente detectar suas dobradiças Fazendo backup de dados e armazenamento que é muito necessário