# GPT Image Hub prompt

## 객체 감지 네트워크 아키텍처 다이어그램
- URL: https://www.gptimagehub.com/ko/library/object-detection-network-architecture-diagram
- Generator URL: https://www.gptimagehub.com/ko/generate?promptId=oss-ym-g2-25063
- Markdown URL: https://www.gptimagehub.com/api/agent/markdown?path=%2Fko%2Flibrary%2Fobject-detection-network-architecture-diagram
- Category: 인포그래픽 / 교육
- Model: gpt-image-2
- Aspect ratio: 16:9
- Style: graphic-design
- Tags: 인포그래픽, 아키텍처 다이어그램, 딥러닝, 신경망, 벡터 일러스트, 컴퓨터 비전
- Updated: 2026-10-27T11:34:58.098Z
- Created: 2026-10-27T11:34:58.098Z

딥러닝 감지 파이프라인을 보여주는 깔끔한 2D 벡터 기술 인포그래픽을 생성합니다. 모델 이름이나 서브 블록을 변경하여 다양한 신경망 구조를 표현할 수 있습니다.

### Full prompt
```text
A clean technical architecture diagram explaining [YOLOv13] as a one-page educational slide, showing how an input image flows through Backbone, HyperACE, Neck, FullPAD, and Head detection outputs.

Canvas and Style: Wide 16:9 landscape infographic on a pure white background, crisp 2D vector graphic style, thin colored outlines, rounded rectangular modules, directional arrow connectors, and crisp, legible sans-serif typography.

Layout:
- Left-to-right horizontal pipeline flow with supplementary modules positioned below the main sequence.
- Far left: A stacked thumbnail labeled "Input Image (3, H, W)" shown as layered landscape images, connecting via arrow into the Backbone panel.
- Upper-left: A large blue-outlined panel titled "Backbone" with subtitle "(built from DS-C3k2 blocks)". Inside are 5 vertically stacked DS-C3k2 boxes labeled B1, B2, B3, B4, B5, with "/2" downsampling markers between adjacent blocks (4 total). B1 and B2 branch right as "B1 (unused)" and "B2 (unused)". B3, B4, and B5 branch right via blue arrows labeled "B3 (C3, H/8, W/8)", "B4 (C4, H/16, W/16)", and "B5 (C5, H/32, W/32)".
- Lower-left / center: An orange-outlined module titled "HyperACE" with subtitle "(Hypergraph-based Adaptive Correlation Enhancement) Eq.8-11". Inside is a dashed rounded note box reading "Models cross-scale and cross-position high-order and low-order correlations". Blue feeds from B3, B4, and B5 enter HyperACE. HyperACE outputs an orange arrow to an orange 3D isometric cube labeled "Y" and "Correlation-enhanced feature".
- Upper-center: A green-outlined panel titled "Neck" with subtitle "(built from DS-C3k2 blocks)". Contains 6 DS-C3k2 block boxes in a 2-column by 3-row grid with connecting arrows, 2 yellow upsample boxes labeled "↑ 2×", 2 purple downsample boxes labeled "↓ 2×", 3 blue circular input junctions on the left edge, and 3 tunnel markers below.
- Lower-center / right: A red-outlined legend module titled "FullPAD" with subtitle "(Full-Pipeline Aggregation-and-Distribution) Eq.12-13". Contains 3 legend rows: "Backbone–Neck tunnel (Backbone-to-Neck connection)" with blue dotted arrows and hollow circles; "In-Neck tunnel (the internal layers of the Neck)" with green dotted arrows and hollow circles; "Neck–Head tunnel (Neck-to-Head connection)" with purple dotted arrows and hollow circles. The cube feature feeds into FullPAD, which routes dotted blue, green, and purple paths to the Neck and Head.
- Far right: A purple-outlined panel titled "Head" with subtitle "(Detect x3)". Contains 3 vertically stacked boxes: "Detect (stride 8)", "Detect (stride 16)", and "Detect (stride 32)", outputting rightward as "H3 (scale 1/8)", "H4 (scale 1/16)", and "H5 (scale 1/32)".

Bottom Caption Box: A wide rounded banner across the bottom. Line 1: "Classic YOLO is a one-directional Backbone → Neck → Head pipeline." Line 2: "[YOLOv13] adds HyperACE and FullPAD on top of it to enhance feature representation and information flow."

Constraints: Perfectly legible text, neat geometric alignment, exact module counts, pastel fills for inner boxes, no extra decorative modules, flat vector rendering.
```