Replies: 1 comment
-
这是由于文本检测模型没有检测到“Кухня”,可适当调低text_det_box_thresh |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
问题描述
我使用 PaddleOCR 对从 SVG 转换而来的楼层平面图(通过


svg_to_image()
)进行 OCR 识别,但是“Кухня”(厨房)的文本没有被识别出来。右侧是识别结果截图:
![识别结果示例]
环境
eslav_PP-OCRv5_mobile_rec
svg_to_image()
函数重现步骤
svg_to_image()
转为 PNG。• “Кухня” 不在 rec_texts 列表中。
• 其他房间名识别正常。
可能原因
• SVG 转 PNG 后文字被抗锯齿或渲染失真。
• 模型对某些西里尔字母变体(例如 “ю”、“х”)识别率不高。
• OCR 参数(阈值、方向分类等)未针对该场景优化。
请求帮助
• 有哪些方法能让 PaddleOCR 更稳定地识别 “Кухня”?
• 转换 SVG 时是否需要调整分辨率、DPI 或抗锯齿设置?
• OCR 初始化参数有哪些建议微调方向?
感谢各位指点!
Problem Description
I’m using PaddleOCR to perform OCR on a floorplan diagram converted from SVG to PNG (via svg_to_image()), but the text “Кухня” (kitchen) is not recognized.
Here is a screenshot of the result next to the original:
Environment
• OS: macOS / Linux
• Python: 3.12
• PaddleOCR version: eslav_PP-OCRv5_mobile_rec
• Other components: ultralytics YOLO (door/window detection), FastAPI
• SVG → PNG conversion: custom svg_to_image() function
Steps to Reproduce
1. Convert the original SVG to PNG using svg_to_image().
2. Initialize PaddleOCR with: Like Chinese 2. python
3. Call ocr.predict(img_path) and inspect rec_texts.
4. You’ll see room names like “Коридор”, “Ванная”, “Спальня”, “Лоджия”, but “Кухня” is missing.
Expected Result
The OCR engine should detect and output: Кухня 12.06
into the rec_texts list.
Actual Result
• “Кухня” does not appear in rec_texts.
• Other Cyrillic room names are correctly recognized.
Possible Causes
• Text rendering artifacts after SVG → PNG conversion (antialiasing, DPI).
• Recognition difficulties for certain Cyrillic character variants (“ю”, “х”).
• OCR parameter thresholds or orientation settings not tuned for this use case.
Help Requested
• Suggestions for improving PaddleOCR’s accuracy on “Кухня”.
• Recommended pre-/post-processing steps (e.g. resizing, sharpening, DPI adjustments).
• Parameter tuning tips (box thresholds, score thresholds, orientation flags).
Thank you very much for your help!
Beta Was this translation helpful? Give feedback.
All reactions