GSP972

Visão geral

A Vertex AI reúne os serviços do Google Cloud para criar machine learning (ML) em uma interface e API unificadas. Na Vertex AI, é muito fácil treinar e comparar modelos usando o AutoML ou o treinamento de código personalizado. Além disso, os modelos são armazenados em um repositório central. Agora é possível implantar esses modelos nos mesmos endpoints da Vertex AI.

O AutoML da Vertex AI ajuda quem ainda não tem muita experiência com ML a treinar modelos de classificação de imagens de alta qualidade. Neste laboratório, você vai aprender a criar um modelo personalizado de ML que identifica automaticamente partes danificadas de um carro. Como o tempo necessário para treinar o modelo é maior que a duração do laboratório, você vai solicitar previsões a um modelo hospedado em outro projeto com o mesmo conjunto de dados. Depois você vai ajustar os valores dos dados na solicitação e ver como isso altera a previsão do modelo.

Objetivos

Neste laboratório, você aprenderá a fazer o seguinte:

• Fazer o upload de um conjunto de dados rotulado para o Cloud Storage usando um arquivo CSV e conectá-lo à Vertex AI como um conjunto de dados gerenciado
• Inspecionar as imagens enviadas para garantir que não há erros no conjunto de dados.
• Iniciar um job de treinamento de modelo do AutoML.
• Solicitar previsões a um modelo hospedado que foi treinado com o mesmo conjunto de dados.

Configuração e requisitos

Antes de clicar no botão Começar o Laboratório

Leia estas instruções. Os laboratórios são cronometrados e não podem ser pausados. O timer é ativado quando você clica em Iniciar laboratório e mostra por quanto tempo os recursos do Google Cloud vão ficar disponíveis.

Este laboratório prático permite que você realize as atividades em um ambiente real de nuvem, e não em uma simulação ou demonstração. Você vai receber novas credenciais temporárias para fazer login e acessar o Google Cloud durante o laboratório.

Confira os requisitos para concluir o laboratório:

• Acesso a um navegador de Internet padrão (recomendamos o Chrome).

Observação: para executar este laboratório, use o modo de navegação anônima (recomendado) ou uma janela anônima do navegador. Isso evita conflitos entre sua conta pessoal e de estudante, o que poderia causar cobranças extras na sua conta pessoal.

• Tempo para concluir o laboratório: não se esqueça que, depois de começar, não será possível pausar o laboratório.

Observação: use apenas a conta de estudante neste laboratório. Se usar outra conta do Google Cloud, você poderá receber cobranças nela.

Como iniciar seu laboratório e fazer login no console do Google Cloud

1. Clique no botão Começar o laboratório. Se for preciso pagar por ele, uma caixa de diálogo vai aparecer para você selecionar a forma de pagamento. No painel Detalhes do Laboratório, à esquerda, você vai encontrar o seguinte:

   • O botão Abrir Console do Google Cloud
   • O tempo restante
   • As credenciais temporárias que você vai usar neste laboratório
   • Outras informações, se forem necessárias

2. Se você estiver usando o navegador Chrome, clique em Abrir console do Google Cloud ou clique com o botão direito do mouse e selecione Abrir link em uma janela anônima.

   O laboratório ativa os recursos e depois abre a página Fazer Login em outra guia.

   Dica: coloque as guias em janelas separadas lado a lado.

   Observação: se aparecer a caixa de diálogo Escolher uma conta, clique em Usar outra conta.

3. Se necessário, copie o Nome de usuário abaixo e cole na caixa de diálogo Fazer login.

   {{{user_0.username | "Username"}}}

   Você também encontra o nome de usuário no painel Detalhes do Laboratório.

4. Clique em Próxima.

5. Copie a Senha abaixo e cole na caixa de diálogo de Olá.

   {{{user_0.password | "Password"}}}

   Você também encontra a senha no painel Detalhes do Laboratório.

6. Clique em Próxima.

   Importante: você precisa usar as credenciais fornecidas no laboratório, e não as da sua conta do Google Cloud. Observação: se você usar sua própria conta do Google Cloud neste laboratório, é possível que receba cobranças adicionais.

7. Acesse as próximas páginas:

   • Aceite os Termos e Condições.
   • Não adicione opções de recuperação nem autenticação de dois fatores (porque essa é uma conta temporária).
   • Não se inscreva em testes gratuitos.

Depois de alguns instantes, o console do Google Cloud será aberto nesta guia.

Observação: para acessar os produtos e serviços do Google Cloud, clique no Menu de navegação ou digite o nome do serviço ou produto no campo Pesquisar.

Ativar o Cloud Shell

O Cloud Shell é uma máquina virtual com várias ferramentas de desenvolvimento. Ele tem um diretório principal permanente de 5 GB e é executado no Google Cloud. O Cloud Shell oferece acesso de linha de comando aos recursos do Google Cloud.

1. Clique em Ativar o Cloud Shell na parte de cima do console do Google Cloud.

2. Clique nas seguintes janelas:

   • Continue na janela de informações do Cloud Shell.
   • Autorize o Cloud Shell a usar suas credenciais para fazer chamadas de APIs do Google Cloud.

Depois de se conectar, você verá que sua conta já está autenticada e que o projeto está configurado com seu Project_ID, . A saída contém uma linha que declara o projeto PROJECT_ID para esta sessão:

Your Cloud Platform project in this session is set to {{{project_0.project_id | "PROJECT_ID"}}}

A gcloud é a ferramenta de linha de comando do Google Cloud. Ela vem pré-instalada no Cloud Shell e aceita preenchimento com tabulação.

3. (Opcional) É possível listar o nome da conta ativa usando este comando:

gcloud auth list

4. Clique em Autorizar.

Saída:

ACTIVE: * ACCOUNT: {{{user_0.username | "ACCOUNT"}}} To set the active account, run: $ gcloud config set account `ACCOUNT`

5. (Opcional) É possível listar o ID do projeto usando este comando:

gcloud config list project

Saída:

[core] project = {{{project_0.project_id | "PROJECT_ID"}}} Observação: consulte a documentação completa da gcloud no Google Cloud no guia de visão geral da gcloud CLI.

Tarefa 1: Fazer o upload das imagens de treinamento para o Cloud Storage

Nesta tarefa, você vai fazer o upload das imagens de treinamento que escolheu usar no Cloud Storage. Assim fica mais fácil importar os dados para a Vertex AI depois.

Para treinar um modelo que classifique imagens de partes danificadas de um carro, é preciso fornecer dados de treinamento rotulados à máquina. O modelo vai usar esses dados para entender cada imagem, reconhecer as partes do carro e indicar as que estão danificadas.

Observação: no laboratório, você não precisa rotular as imagens porque vai usar um arquivo CSV com um conjunto de dados já rotulado (imagem e rótulo). Na próxima seção, vamos descrever as etapas para usar o arquivo CSV.

Neste exemplo, o modelo vai aprender a classificar cinco partes danificadas de um carro: para-choque, compartimento do motor, capô, lataria e para-brisa.

Crie um bucket do Cloud Storage

• Abra uma janela do Cloud Shell e execute o seguinte comando para criar um bucket do Cloud Storage:

gsutil mb -p {{{project_0.project_id | PROJECT_ID}}} \ -c standard \ -l "{{{project_0.default_region | REGION}}}" \ gs://{{{project_0.project_id | BUCKET}}}

Faça o upload das imagens de carros no bucket do Storage

As imagens de treinamento estão disponíveis em um bucket público do Cloud Storage. Mais uma vez, copie e cole o modelo de script abaixo no Cloud Shell para copiar as imagens para o seu bucket.

1. Para salvar as imagens no bucket do Cloud Storage, execute este comando:

gsutil -m cp -r gs://car_damage_lab_images/* gs://{{{project_0.project_id | BUCKET}}}

2. No painel de navegação, clique em Cloud Storage > Buckets.

3. Clique no botão Atualizar na parte de cima do navegador do Cloud Storage.

4. Clique no nome do bucket. Ele deve ter cinco pastas de fotos, uma para cada parte danificada a ser classificada:

5. Outra opção é clicar em uma das pastas para conferir as imagens nela.

Ótimo. As imagens de carros estão organizadas e prontas para o treinamento.

Clique em Verificar meu progresso para conferir o objetivo. Faça o upload das imagens de carros no bucket do Storage

Tarefa 2: Criar um conjunto de dados

Agora você vai criar um conjunto de dados e conectá-lo às imagens de treinamento para que a Vertex AI tenha acesso a elas.

Normalmente, você criaria um arquivo CSV com um URL em cada linha para as imagens de treinamento e os rótulos associados a elas. Como o arquivo CSV já foi criado, você só precisa alterar o nome do bucket nele e fazer upload do arquivo para o bucket do Cloud Storage.

Atualize o arquivo CSV

Copie e cole os modelos de script abaixo no Cloud Shell e pressione "Enter" para atualizar o arquivo. Depois faça o upload do CSV.

1. Para criar uma cópia do arquivo, execute este comando:

gsutil cp gs://car_damage_lab_metadata/data.csv .

2. Para atualizar o CSV com o caminho para o armazenamento, execute este comando:

sed -i -e "s/car_damage_lab_images/{{{project_0.project_id | BUCKET}}}/g" ./data.csv

3. Execute o seguinte comando para verificar se o nome do bucket foi inserido corretamente no CSV:

cat ./data.csv

4. Para fazer upload do arquivo CSV no bucket do Cloud Storage, execute este comando:

gsutil cp ./data.csv gs://{{{project_0.project_id | BUCKET}}}

5. Depois que o comando for executado, clique no botão Atualizar na parte de cima do navegador do Cloud Storage e abra o bucket.

6. Confirme que o arquivo data.csv está no bucket.

Ativar o Gemini Code Assist no Cloud Shell IDE

Você pode usar o Gemini Code Assist em um ambiente de desenvolvimento integrado (IDE), como o Cloud Shell, para receber orientações sobre o código ou resolver problemas com ele. Antes de começar a usar o Gemini Code Assist, você precisa ativá-lo.

1. No Cloud Shell, ative a API Gemini para Google Cloud com o seguinte comando:

gcloud services enable cloudaicompanion.googleapis.com

2. Clique em Abrir editor na barra de ferramentas do Cloud Shell.

Observação: para abrir o editor do Cloud Shell, clique em Abrir editor na barra de ferramentas do Cloud Shell. Você pode alternar o Cloud Shell e o editor de código clicando em Abrir editor ou Abrir terminal, conforme necessário.

3. No painel à esquerda, clique no ícone Configurações e, na visualização Configurações, pesquise Gemini Code Assist.

4. Localize e verifique se a caixa de seleção Geminicodeassist: Ativar está marcada e feche as Configurações.

5. Clique em Cloud Code - Sem projeto na barra de status na parte de baixo da tela.

6. Autorize o plug-in conforme instruído. Se um projeto não for selecionado automaticamente, clique em Selecionar um projeto do Google Cloud e escolha .

7. Verifique se o projeto do Google Cloud () aparece na mensagem de status do Cloud Code na barra de status.

Analise o arquivo CSV

Para aumentar a produtividade e minimizar a troca de contexto, o Gemini Code Assist oferece ações inteligentes com tecnologia de IA diretamente no editor de código. Nesta seção, você vai pedir ao Gemini Code Assist para explicar a estrutura e a finalidade do arquivo CSV para um membro da equipe.

1. No editor do Cloud Shell, abra o explorador de arquivos e o arquivo data.csv. Essa ação ativa o Gemini Code Assist, conforme indicado pela presença do ícone no canto superior direito do editor.

2. Clique no ícone Gemini Code Assist: Smart Actions e selecione Explain this.

3. O Gemini Code Assist abre um painel de chat com o comando Explicar isto preenchido. Na caixa de texto inline do chat do Code Assist, substitua o comando preenchido previamente pelo seguinte e clique em Enviar:

Você é engenheiro de machine learning na Cymbal AI. Forneça uma análise formal dos dados CSV data.csv, que é um manifesto de treinamento para um modelo de classificação de imagens de rótulo único da Vertex AI. Explique a estrutura e a finalidade do arquivo para ajudar um novo membro da equipe a entender esse conjunto de dados, que foi projetado para reconhecer peças de carros danificadas. Com base nesses dados, forneça os seguintes insights em um formato estruturado e profissional: * Visão geral do conjunto de dados: descreva a finalidade e a estrutura das três colunas (divisão do conjunto de dados, URI da imagem e rótulo). * Divisões do conjunto de dados: forneça uma contagem do número total de imagens nas divisões de TREINAMENTO, VALIDAÇÃO e TESTE. * Distribuição de classes: para cada classe (bumper, engine_compartment, hood, lateral, windshield), forneça uma contagem do número de imagens atribuídas a ela. * Análise de desequilíbrio de dados: com base nas contagens de classe e divisão, avalie se o conjunto de dados é equilibrado ou desequilibrado. Explique como isso pode afetar o desempenho do modelo durante o treinamento. Para as melhorias sugeridas, não faça alterações no conteúdo do arquivo.

Explicações detalhadas sobre a estrutura e a finalidade do código do arquivo CSV data.csv aparecem no chat do Gemini Code Assist.

Crie um conjunto de dados gerenciado

1. No console do Google Cloud, abra o menu de navegação () e clique em Vertex AI > Painel.

2. Selecione Ativar todas as APIs recomendadas se ainda não tiver feito isso.

3. No menu de navegação à esquerda da Vertex AI, clique em Conjuntos de dados.

4. Clique em + Criar na parte de cima do console.

5. No campo "Nome do conjunto de dados", digite damaged_car_parts.

6. Selecione Classificação de rótulo único. Se você fizer classificação multiclasse nos seus próprios projetos, escolha a opção "Classificação de vários rótulos".

7. Selecione a região como .

8. Clique em Criar.

Conecte o conjunto de dados às imagens de treinamento

Nesta seção, você vai escolher um local para as imagens de treinamento que enviou por upload na etapa anterior.

1. Na seção Selecionar um método de importação, clique em Selecione os arquivos de importação do Cloud Storage.

2. Na seção Selecione os arquivos de importação do Cloud Storage, clique em Procurar.

3. Siga as instruções para navegar até o bucket do Storage que você criou e clique no arquivo data.csv. Clique em Selecionar.

4. Depois de você selecionar o arquivo, uma caixa de seleção verde vai aparecer à esquerda do caminho do arquivo. Clique em Continuar para prosseguir.

Observação: pode levar de 9 a 12 minutos para que as imagens sejam importadas e alinhadas às respectivas categorias. Aguarde essa etapa ser concluída para conferir seu progresso.

5. Quando a importação estiver concluída, é só clicar na guia Procurar para se preparar para a próxima seção. Dica: talvez seja necessário atualizar a página para confirmar.

Clique em Verificar meu progresso para conferir o objetivo. Crie um conjunto de dados

Tarefa 3: Inspecionar as imagens

Agora você vai analisar as imagens para garantir que não há erros no conjunto de dados.

Verifique os rótulos das imagens

1. Depois de atualizar a página do navegador, clique em Conjuntos de dados, selecione o nome da imagem e clique em Procurar.

2. Em Filtrar rótulos, clique em um rótulo para ver as imagens de treinamento específicas (por exemplo, engine_compartment).

Observação: se quiser criar um modelo de produção, serão necessárias pelo menos cem imagens por rótulo para garantir alta acurácia. Como esta é apenas uma demonstração, foram usadas somente 20 imagens de cada tipo para agilizar o treinamento do modelo.

3. Caso uma imagem seja rotulada incorretamente, clique nela para selecionar o rótulo correto ou excluí-la do conjunto de treinamento:

4. Depois clique na guia Analisar para saber quantas imagens cada rótulo tem. A janela Estatísticas do rótulo vai aparecer no navegador.

Tarefa 4: Treinar o modelo

Agora você pode começar a treinar o modelo. A Vertex AI cuida disso automaticamente, para você não precisar escrever o código do modelo.

1. Clique em Treinar novo modelo à direita.

2. Na janela Método de treinamento, mantenha as configurações padrão e selecione AutoML como método de treinamento. Clique em Continuar.

3. Na janela Detalhes do modelo, defina o nome do modelo como damaged_car_parts_model. Clique em Continuar.

4. Na janela Opções de treinamento, selecione Ativar treinamento incremental e clique em Continuar.

5. Na janela Computação e preços, defina seu orçamento como no máximo 8 horas de uso de nós.

6. Clique em Iniciar treinamento.

Observação: talvez o treinamento do modelo demore mais do que a duração do laboratório. Você não precisa esperar o treinamento do modelo terminar para prosseguir para a próxima seção.

Clique em Verificar meu progresso para conferir o objetivo. Treine o modelo

Tarefa 5: Solicitar uma previsão do modelo hospedado

É provável que o treinamento do modelo local demore mais que a duração do laboratório. Por isso, há um modelo hospedado em outro projeto, treinado com o mesmo conjunto de dados, para você solicitar previsões enquanto o treinamento local está em andamento.

Já criamos um proxy para o modelo pré-treinado, então você não precisa fazer mais nada para trabalhar no ambiente do laboratório.

Para solicitar previsões, você as envia a um endpoint dentro do seu projeto, que encaminha a solicitação ao modelo hospedado e retorna a resposta. A forma de enviar previsões ao proxy do AutoML é muito semelhante à forma de interagir com o modelo que você criou, então esta é uma oportunidade para praticar.

Veja o nome do endpoint do proxy do AutoML

1. No menu de navegação (≡) do console do Google Cloud, clique em Cloud Run.

2. Clique em automl-proxy.

3. Copie o URL do endpoint. Ele tem este formato: https://automl-proxy-xfpm6c62ta-uc.a.run.app.

Você vai usar o endpoint para solicitar a previsão na seção seguinte.

Crie uma solicitação de previsão

1. Abra uma nova janela do Cloud Shell.

2. Na barra de ferramentas do Cloud Shell, clique em Abrir editor. Se solicitado, clique em Abrir em uma nova janela.

3. Clique em Arquivo > Novo arquivo.

4. Insira o nome do arquivo como payload.json na janela Novo arquivo e selecione o caminho no menu suspenso (/home/student_xx_xxxxx).

5. Clique em OK.

6. Cole o código a seguir no arquivo que você criou:

{ "instances": [{ "content": "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" }], "parameters": { "confidenceThreshold": 0.5, "maxPredictions": 5 } }

7. Clique em Arquivo > Salvar.

O código que você adicionou ao arquivo é uma string Base64 da imagem a seguir.

Agora, você decide pedir ajuda ao Gemini Code Assist para explicar a estrutura e a finalidade do payload JSON a um membro da equipe.

8. Com o arquivo payload.json aberto, clique no ícone Gemini Code Assist: Ações inteligentes na barra de ferramentas e selecione Explain this.

9. O Gemini Code Assist abre um painel de chat com o comando Explain this preenchido. Na caixa de texto inline do chat do Code Assist, substitua o comando preenchido previamente pelo seguinte e clique em Enviar:

Você é engenheiro de machine learning na Cymbal AI. Um novo membro da equipe precisa de ajuda para entender a estrutura e a finalidade do arquivo payload.json. Esse payload é usado para solicitar uma previsão de um modelo de classificação de imagens da Vertex AI implantado. Sua explicação deve abordar: * A finalidade geral do payload no contexto de um fluxo de trabalho de previsão da Vertex AI. * O papel da chave "instances", explicando os dados de imagem codificados em base64 nela. * A função da chave "parameters", incluindo a finalidade de "confidenceThreshold" e "maxPredictions". * Como esse payload seria enviado ao endpoint do modelo. Para as melhorias sugeridas, não faça alterações no conteúdo do arquivo.

Uma explicação detalhada do arquivo de payload JSON payload.json aparece no chat do Gemini Code Assist.

10. Agora defina as variáveis de ambiente a seguir. Copie o URL do proxy do AutoML, que você anotou antes.

AUTOML_PROXY=<automl-proxy url> INPUT_DATA_FILE=payload.json

11. Execute o comando a seguir para enviar uma solicitação de API ao endpoint do proxy do AutoML para receber a previsão do modelo hospedado:

curl -X POST -H "Content-Type: application/json" $AUTOML_PROXY/v1 -d "@${INPUT_DATA_FILE}"

Se a previsão tiver funcionado, a resposta deverá ter este formato:

Saída:

{"predictions":[{"confidences":[0.951557755],"displayNames":["bumper"],"ids":["1960986684719890432"]}],"deployedModelId":"4271461936421404672","model":"projects/1030115194620/locations/"{{{project_0.default_region | REGION}}}"/models/2143634257791156224","modelDisplayName":"damaged_car_parts_vertex","modelVersionId":"1"}

Os resultados da previsão do modelo dispensam explicações. O campo displayNames deve prever corretamente bumper com alto limite de confiança. Agora você pode mudar o valor da imagem em código Base64 no arquivo JSON criado.

Clique em Verificar meu progresso para conferir o objetivo. Crie a solicitação de previsão

12. Clique com o botão direito do mouse em cada uma das imagens abaixo e selecione Salvar imagem como….

13. Siga as instruções para salvar cada imagem com um nome exclusivo. Dica: use nomes simples, como "Imagem1" e "Imagem2", para facilitar o upload mais tarde.

14. Abra o Base64 Image Encoder e siga as instruções para fazer upload e codificar uma imagem em uma string Base64.

15. Substitua o valor da string Base64 no campo content do arquivo de payload JSON e execute a previsão de novo. Repita o procedimento com as demais imagens.

Como o modelo se saiu? Ele fez a previsão das três imagens corretamente? As respostas precisam ser estas, respectivamente:

{"predictions":[{"ids":["5419751198540431360"],"confidences":[0.985487759],"displayNames":["engine_compartment"]}],"deployedModelId":"4271461936421404672","model":"projects/1030115194620/locations/"{{{project_0.default_region | REGION}}}"/models/2143634257791156224","modelDisplayName":"damaged_car_parts_vertex","modelVersionId":"1"} {"predictions":[{"displayNames":["hood"],"ids":["3113908189326737408"],"confidences":[0.962432086]}],"deployedModelId":"4271461936421404672","model":"projects/1030115194620/locations/"{{{project_0.default_region | REGION}}}"/models/2143634257791156224","modelDisplayName":"damaged_car_parts_vertex","modelVersionId":"1"}

Parabéns!

Neste laboratório, você aprendeu a treinar um modelo personalizado de machine learning com a ajuda do Gemini Code Assist e a gerar previsões com um modelo hospedado usando solicitações de API. Você enviou as imagens de treinamento por upload ao Cloud Storage e usou um arquivo CSV na Vertex AI para encontrá-las. Depois inspecionou as imagens rotuladas para encontrar possíveis discrepâncias antes de avaliar o modelo treinado. Agora você já sabe treinar um modelo com seu próprio conjunto de dados de imagens.

Próximas etapas / Saiba mais

• Leia a documentação da Vertex AI para saber mais.

Treinamento e certificação do Google Cloud

Esses treinamentos ajudam você a aproveitar as tecnologias do Google Cloud ao máximo. Nossas aulas incluem habilidades técnicas e práticas recomendadas para ajudar você a alcançar rapidamente o nível esperado e continuar sua jornada de aprendizado. Oferecemos treinamentos que vão do nível básico ao avançado, com opções de aulas virtuais, sob demanda e por meio de transmissões ao vivo para que você possa encaixá-las na correria do seu dia a dia. As certificações validam sua experiência e comprovam suas habilidades com as tecnologias do Google Cloud.

Manual atualizado em 15 de outubro de 2025

Laboratório testado em 10 de setembro de 2025

Copyright 2025 Google LLC. Todos os direitos reservados. Google e o logotipo do Google são marcas registradas da Google LLC. Todos os outros nomes de produtos e empresas podem ser marcas registradas das respectivas empresas a que estão associados.