プロジェクトを作成する
ワークスペースにプロジェクトを作成します。
概要
モデルをトレーニングする前に、Project を作成する必要があります。
Project には画像または動画と、そのアノテーションが含まれます。このデータは、その後、データセットのバージョン、つまり時間の中で固定されたデータのスナップショットに変換できます。バージョンは、その後モデルのトレーニングに使用できます。
Webアプリ
Project を作成する
まず、Roboflow ダッシュボードに移動します。次に、「Create New Project」をクリックします:

新しいプロジェクトを作成できるページに移動します:

このページでは、以下を入力する必要があります:
プロジェクトの種類.
Object Detection: 画像内のオブジェクトの位置を見つけます。
Single-Label Classification: 限られたカテゴリの中から、画像にラベルを 1 つ割り当てます。
Multi-Label Classification: 限られたカテゴリの中から、画像に関連する任意の数のラベルを割り当てます。
Instance Segmentation: ピクセル単位で、画像内のオブジェクトの位置を見つけます。
Action Recognition: 箱を開けるなど、動画内の動作とその発生時刻を識別します。
Semantic Segmentation: ピクセル単位で、画像内のオブジェクトの位置を見つけ、見つかった各オブジェクトに固有の参照を作成します。
Keypoint Detection: 画像内のオブジェクトとそのキーポイントの位置を見つけます。一般的には、オブジェクトの姿勢を判定するために使用されます。
プロジェクト名: プロジェクトの名前。
アノテーショングループ: 画像内で何を検出しているかを分類するラベルです(例: 「chess pieces」「vehicles」「defects」)。同じアノテーショングループを共有するプロジェクトは、クラス一覧とアノテーションも共有します。詳細は Annotation Groups をご覧ください。
これらの値を指定したら、フォームを送信してプロジェクトを作成します。
別の種類のサポート対象プロジェクトを見たい場合は、プロジェクト種類のドロップダウンからそのオプションを選択して関心を示すことができます。
Agent から Project を作成する
から Project を作成できます Roboflow Agent チャットを離れずに。"+" タブを開き、「Create a New Asset」ページの「New Model」をクリックします。プロジェクト種類セレクターと他のフィールドを含む同じフォームが、タブ内に開きます。
Project が作成されると、そのタブはその Project のビューに切り替わります。
HTTP API
プロジェクトは空の状態で始まります。次に、 データを追加 します。フォルダ内にプロジェクトを作成するには、 Project Folders からフォルダ ID を group.
Creates an empty project. Requires project:create. The URL slug is made from name and gets a random suffix when the slug is taken.
Roboflow API key or OAuth access token.
Workspace URL slug.
Display name. Letters, numbers, dashes, underscores, and spaces only.
Project type. The response reports both classification types as classification. anomaly-detection needs Anomaly Detection turned on for your workspace, creates a single-label classification project with the fixed classes normal and anomalous, and needs an annotation group that no other project type uses.
Annotation group: a noun for what you label (ex: sharks, defects). Projects in the same group share classes.
Private makes a private project, if your plan allows it. Any other value makes a public project on Roboflow Universe. Leave out for a private project. Required on public plans, which cannot create private projects.
ID of the project folder to create the project in.
The new project.
Workspace and project URL slugs joined by a slash.
Project type (ex: object-detection, classification, instance-segmentation).
Project display name.
Creation time in seconds since the Unix epoch.
Last update time in seconds since the Unix epoch.
Images in the project.
Images with no annotations.
Annotation group name.
Dataset versions generated from the project.
True when the project is public on Roboflow Universe.
True for multi-label classification projects.
Dataset license (ex: Private, CC BY 4.0).
Thumbnail shown for the project, or null when it has none.
Your plan does not allow private projects, or an anomaly-detection project cannot be created (the feature is off, or the annotation group is used by another project type).
Authentication is missing or invalid.
Public projects are turned off for this workspace.
The resource does not exist or is not accessible.
group is not an existing project folder.
name, type, annotation, or license is missing or invalid, or the workspace is not set up yet.
The request rate is too high.
POST /{workspace}/projects HTTP/1.1
Host: api.roboflow.com
Authorization: Bearer YOUR_SECRET_TOKEN
Content-Type: application/json
Accept: */*
Content-Length: 93
{
"name": "Sharks Dataset",
"type": "object-detection",
"annotation": "sharks",
"license": "Private"
}{
"id": "my-workspace/sharks-dataset",
"type": "object-detection",
"name": "Sharks Dataset",
"created": 1688739471.567,
"updated": 1688739471.567,
"images": 0,
"unannotated": 0,
"annotation": "sharks",
"versions": 0,
"public": false,
"multilabel": false,
"license": "Private",
"splits": {},
"colors": {},
"classes": {}
}Python SDK
Workspace.create_project() はワークスペースに新しいプロジェクトを作成し、その後画像をアップロードできる プロジェクト オブジェクトを返します。
パラメータ
project_name(str) - 表示名です。URL のスラッグはこれから自動生成されます。project_type(str) - 以下のいずれかです:object-detectionsingle-label-classificationmulti-label-classificationinstance-segmentationsemantic-segmentationkeypoint-detection
project_license(str) - 以下に設定します"Private"プライベートプロジェクトの場合。公開ライセンスの値には"MIT","CC BY 4.0","Public Domain"などがあります。完全な一覧は Web アプリのプロジェクト作成フォームを参照してください。annotation(str) - アノテーショングループ: 何にラベルを付けるかを表す名詞("flowers","vehicles","defects")。ラベリング UI のプロンプトで使用されます。
CLI
コマンドラインから新しいプロジェクトを作成できます。
コマンド
オプション
--type
プロジェクト種類(必須)。対応している種類は以下を参照してください
--license
プロジェクトのライセンス(任意)
--annotation
アノテーショングループ名(任意、既定値はプロジェクト名)
対応するプロジェクト種類
object-detectionsingle-label-classificationmulti-label-classificationinstance-segmentationsemantic-segmentationkeypoint-detection
例
オブジェクト検出プロジェクトを作成します:
分類プロジェクトを作成します:
JSON出力
MCPサーバー
AIエージェントを MCPサーバー また、これらのツールでプロジェクトを作成および確認できます:
projects_create
新しいコンピュータビジョンプロジェクトを作成します。
projects_get
バージョン、クラス、分割、学習済みモデルを含むプロジェクトの詳細を取得します。
projects_list
ワークスペース内のプロジェクトを一覧表示します。
次のステップ
画像やその他のデータをプロジェクトに追加します。詳細は データを追加.
モデルのトレーニングに使えるように画像へラベルを付けます。詳細は Roboflow Annotate の概要.
最終更新
役に立ちましたか?