Mit Sammlungen den Überblick behalten
Sie können Inhalte basierend auf Ihren Einstellungen speichern und kategorisieren.
Auf dieser Seite finden Sie Voraussetzungen und eine detaillierte Anleitung zur Feinabstimmung von Gemini mithilfe von Bilddaten und beaufsichtigtem Lernen.
Anwendungsfälle
Mit der Feinabstimmung können Sie Gemini-Basismodelle für spezielle Aufgaben anpassen.
Hier einige Anwendungsfälle für Bilder:
Produktkatalog-Optimierung: Es werden wichtige Attribute aus Bildern extrahiert (z.B. Marke, Farbe, Größe), um Ihren Produktkatalog automatisch zu erstellen und zu ergänzen.
Bildmoderation: Sie können ein Modell optimieren, um unangemessene oder schädliche Inhalte in Bildern zu erkennen und zu melden. So wird das Internet für alle sicherer.
Visuelle Prüfung: Trainieren Sie ein Modell, um bestimmte Objekte oder Mängel in Bildern zu erkennen und so die Qualitätskontrolle oder Prüfprozesse zu automatisieren.
Bildklassifizierung: Verbessern Sie die Genauigkeit der Bildklassifizierung für bestimmte Bereiche wie die medizinische Bildgebung oder die Analyse von Satellitenbildern.
Bildbasierte Empfehlungen: Bilder werden analysiert, um personalisierte Empfehlungen zu geben, z. B. ähnliche Produkte oder ergänzende Artikel.
Extraktion von Tabelleninhalten: Daten aus Tabellen in Bildern extrahieren und in strukturierte Formate wie Tabellenkalkulationen oder Datenbanken konvertieren.
Beschränkungen
Maximale Anzahl von Bildern pro Beispiel: 30
Maximale Größe der Bilddatei: 20 MB
Weitere Informationen zu den Anforderungen an Bildbeispiele finden Sie auf der Seite Bildverständnis.
Dataset-Format
Der fileUri für Ihr Dataset kann der URI für eine Datei in einem Cloud Storage-Bucket oder eine öffentlich zugängliche HTTP- oder HTTPS-URL sein.
Im Folgenden finden Sie ein Beispiel für ein Bild-Dataset.
{"contents":[{"role":"user","parts":[{"fileData":{"mimeType":"image/jpeg","fileUri":"gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/image/longcap100/100.jpeg"}},{"text":"Describe this image in detail that captures the essence of it."}]},{"role":"model","parts":[{"text":"A man stands on a road, wearing a blue denim jacket, tan pants, and white sneakers. He has his hands in his pockets and is wearing a white t-shirt under his jacket. The man's pants are cuffed, and his shoes are white. The road is dark grey, and the leaves are green. The man is standing in the shade, and the light is shining on the ground."}]}]}
Beispieldatensätze
Anhand der folgenden Beispiel-Datasets können Sie lernen, wie Sie ein Gemini-Modell abstimmen. Geben Sie die URIs in den entsprechenden Parametern beim Erstellen eines überwachten Textmodell-Feinabstimmungsjobs an, um diese Datasets zu verwenden.
Wenn Sie das Beispiel-Abstimmungs-Dataset verwenden möchten, geben Sie seinen Speicherort so an:
Informationen dazu, wie die überwachte Feinabstimmung in einer Lösung verwendet werden kann, die eine Wissensdatenbank für generative KI erstellt, finden Sie unter Schnellstartlösung: Wissensdatenbank für generative KI.
[[["Leicht verständlich","easyToUnderstand","thumb-up"],["Mein Problem wurde gelöst","solvedMyProblem","thumb-up"],["Sonstiges","otherUp","thumb-up"]],[["Schwer verständlich","hardToUnderstand","thumb-down"],["Informationen oder Beispielcode falsch","incorrectInformationOrSampleCode","thumb-down"],["Benötigte Informationen/Beispiele nicht gefunden","missingTheInformationSamplesINeed","thumb-down"],["Problem mit der Übersetzung","translationIssue","thumb-down"],["Sonstiges","otherDown","thumb-down"]],["Zuletzt aktualisiert: 2025-08-25 (UTC)."],[],[],null,["# Image tuning\n\nThis page provides prerequisites and detailed instructions for fine-tuning\nGemini on image data using supervised learning.\n\nUse cases\n---------\n\nFine-tuning lets you adapt base Gemini models for specialized tasks.\nHere are some image use cases:\n\n- **Product catalog enhancement**: Extract key attributes from images (e.g., brand, color, size) to automatically build and enrich your product catalog.\n- **Image moderation**: Fine-tune a model to detect and flag inappropriate or harmful content in images, ensuring a safer online experience.\n- **Visual inspection**: Train a model to identify specific objects or defects within images, automating quality control or inspection processes.\n- **Image classification**: Improve the accuracy of image classification for specific domains, such as medical imaging or satellite imagery analysis.\n- **Image-based recommendations**: Analyze images to provide personalized recommendations, such as suggesting similar products or complementary items.\n- **Table content extraction**: Extract data from tables within images and convert it into structured formats like spreadsheets or databases.\n\nLimitations\n-----------\n\n- Maximum images per example: 30\n- Maximum image file size: 20MB\n\nTo learn more about image sample requirements, see the [Image understanding](/vertex-ai/generative-ai/docs/multimodal/image-understanding#image-requirements) page.\n\nDataset format\n--------------\n\nThe `fileUri` for your dataset can be the URI for a file in a Cloud Storage\nbucket, or it can be a publicly available HTTP or HTTPS URL.\n\nTo see the generic format example, see\n[Dataset example for Gemini](/vertex-ai/generative-ai/docs/models/gemini-supervised-tuning-prepare#dataset-example).\n\nThe following is an example of an image dataset. \n\n {\n \"contents\": [\n {\n \"role\": \"user\",\n \"parts\": [\n {\n \"fileData\": {\n \"mimeType\": \"image/jpeg\",\n \"fileUri\": \"gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/image/longcap100/100.jpeg\"\n }\n }, \n {\n \"text\": \"Describe this image in detail that captures the essence of it.\"\n }\n ]\n }, \n {\n \"role\": \"model\",\n \"parts\": [\n {\n \"text\": \"A man stands on a road, wearing a blue denim jacket, tan pants, and white sneakers. He has his hands in his pockets and is wearing a white t-shirt under his jacket. The man's pants are cuffed, and his shoes are white. The road is dark grey, and the leaves are green. The man is standing in the shade, and the light is shining on the ground.\"\n }\n ]\n }\n ]\n }\n\n### Sample datasets\n\nYou can use the following sample datasets to learn how to tune a\nGemini model. To use these datasets, specify the URIs in the\napplicable parameters when creating a text model supervised fine-tuning job.\n\nTo use the sample tuning dataset, specify its location as follows: \n\n \"training_dataset_uri\": \"gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/text/sft_train_data.jsonl\",\n\nTo use the sample validation dataset, specify its location as follows: \n\n \"validation_dataset_uri\": \"gs://cloud-samples-data/ai-platform/generative_ai/gemini-2_0/text/sft_validation_data.jsonl\",\n\nWhat's next\n-----------\n\n- To learn more about the image understanding capability of Gemini, see our [Image understanding](/vertex-ai/generative-ai/docs/multimodal/image-understanding) documentation.\n- To start tuning, see [Tune Gemini models by using supervised fine-tuning](/vertex-ai/generative-ai/docs/models/gemini-use-supervised-tuning)\n- To learn how supervised fine-tuning can be used in a solution that builds a generative AI knowledge base, see [Jump Start Solution: Generative AI\n knowledge base](/architecture/ai-ml/generative-ai-knowledge-base)."]]