> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/roboflow/roboflow-ko/train/training-resolutions-by-model-type.md).

# 모델 유형별 학습 해상도

학습 해상도는 모델 정확도, 추론 속도, 학습 시간에 영향을 줍니다. 각 모델 아키텍처에는 이러한 요소의 균형을 맞추는 기본 해상도가 있습니다. 기본적으로 Roboflow는 선택한 모델 아키텍처에 대한 기본 학습 해상도를 제안합니다.

아래 표는 각 모델 아키텍처와 크기에 대한 기본 학습 해상도를 보여줍니다. 새로 [데이터셋 버전](/roboflow/roboflow-ko/datasets/dataset-versions.md).

### 객체 탐지

<table><thead><tr><th>모델 유형</th><th width="273.3359375">패밀리 및 크기</th><th>기본 학습 해상도</th></tr></thead><tbody><tr><td>객체 탐지</td><td>RF-DETR Nano</td><td>384×384</td></tr><tr><td>객체 탐지</td><td>RF-DETR Small</td><td>512×512</td></tr><tr><td>객체 탐지</td><td>RF-DETR Medium</td><td>576×576</td></tr><tr><td>객체 탐지</td><td>RF-DETR Large</td><td>704×704</td></tr><tr><td>객체 탐지</td><td>RF-DETR X Large</td><td>700x700</td></tr><tr><td>객체 탐지</td><td>RF-DETR 2X Large</td><td>880x880</td></tr><tr><td>객체 탐지</td><td>Roboflow 3.0 - Fast</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>Roboflow 3.0 - Accurate</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>Roboflow 3.0 - Medium</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>Roboflow 3.0 - Large</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>Roboflow 3.0 - Extra Large</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv26(n/s/m/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv12 (n/s/m/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv11 (n/s/m/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv10 (n/s/m/b/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv9 (s/m/c/e)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv8 (n/s/m/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv5 (n/s/m/l/x)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLOv7 (legacy)</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLO‑NAS Small</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>YOLO‑NAS Medium</td><td>640×640</td></tr><tr><td>객체 탐지</td><td>Roboflow Instant</td><td>1008x1008</td></tr></tbody></table>

### 인스턴스 세그멘테이션

<table><thead><tr><th>모델 유형</th><th width="272.8203125">패밀리 및 크기</th><th>기본 학습 해상도</th></tr></thead><tbody><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg Nano</td><td>312x312</td></tr><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg Small</td><td>384x384</td></tr><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg Medium</td><td>432x432</td></tr><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg Large</td><td>504x504</td></tr><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg X Large</td><td>624x624</td></tr><tr><td>인스턴스 세그멘테이션</td><td>RF-DETR Seg 2X Large</td><td>768x768</td></tr><tr><td>인스턴스 세그멘테이션</td><td>Roboflow 3.0 - Fast (Seg)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>Roboflow 3.0 - Accurate (Seg)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>Roboflow 3.0 - Medium (Seg)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>Roboflow 3.0 - Large (Seg)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>Roboflow 3.0 - Extra Large (Seg)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>YOLO-seg (v8/10/11/12)</td><td>640×640</td></tr><tr><td>인스턴스 세그멘테이션</td><td>SAM3 (Segment Anything 3)</td><td>1008x1008</td></tr><tr><td>인스턴스 세그멘테이션</td><td>시맨틱 세그멘테이션 (DeepLabV3+)</td><td>≥ 512×512</td></tr></tbody></table>

### 시맨틱 세그멘테이션

<table><thead><tr><th>모델 유형</th><th width="272.8203125">패밀리 및 크기</th><th>기본 학습 해상도</th></tr></thead><tbody><tr><td>시맨틱 세그멘테이션</td><td>YOLO26-SEM (n/s/m/l/x)</td><td>1024×1024</td></tr><tr><td>시맨틱 세그멘테이션</td><td>DeepLabV3+</td><td>≥ 512×512</td></tr></tbody></table>

### 분류 및 포즈

<table><thead><tr><th>모델 유형</th><th width="272.66796875">패밀리 및 크기</th><th>기본 학습 해상도</th></tr></thead><tbody><tr><td>분류 및 포즈</td><td>Resnet-18/34/50</td><td>224x224</td></tr><tr><td>분류 및 포즈</td><td>YOLO-cls (v8/11)</td><td>224x224</td></tr><tr><td>분류 및 포즈</td><td>Vision Transformer (ViT)</td><td>224x224</td></tr><tr><td>분류 및 포즈</td><td>YOLO-pose (키포인트)</td><td>640x640</td></tr><tr><td>키포인트 탐지</td><td>RF-DETR Keypoint (미리보기)</td><td>576x576</td></tr></tbody></table>

### 멀티모달/VLM

<table><thead><tr><th>모델 유형</th><th width="272.96484375">패밀리 및 크기</th><th>기본 학습 해상도</th></tr></thead><tbody><tr><td>멀티모달/VLM</td><td>PaliGemma 2 - 3 B</td><td>448x448</td></tr><tr><td>멀티모달/VLM</td><td>PaliGemma 2 - 10 B/28 B</td><td>448x448</td></tr><tr><td>멀티모달/VLM</td><td>Florence-2</td><td>448x448</td></tr><tr><td>멀티모달/VLM</td><td>QWEN 2.5 VL</td><td>448x448</td></tr><tr><td>멀티모달/VLM</td><td>SmolVLM2</td><td>384x384</td></tr></tbody></table>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.roboflow.com/roboflow/roboflow-ko/train/training-resolutions-by-model-type.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.