PaddlePaddle
triton部署的paddleocr文本检测模型，有置信度输出么？怎么设置
triton部署的paddleocr文本检测模型，有置信度输出么？怎么设置
yaml如下： pipeline_name: OCR
text_type: general
use_doc_preprocessor: False use_textline_orientation: False
SubPipelines: DocPreprocessor: pipeline_name: doc_preprocessor use_doc_orientation_classify: False use_doc_unwarping: False SubModules: DocOrientationClassify: module_name: doc_text_orientation model_name: PP-LCNet_x1_0_doc_ori model_dir: null DocUnwarping: module_name: image_unwarping model_name: UVDoc model_dir: null
SubModules:
TextDetection:
module_name: text_detection
model_name: PP-OCRv5_server_det
model_dir: /app/server/paddleocr_v5/official_models/PP-OCRv5_server_det
limit_side_len: 416
limit_type: min
unclip_ratio: 1.6
TextLineOrientation:
module_name: textline_orientation
model_name: PP-LCNet_x1_0_textline_ori
model_dir: null
batch_size: 6
TextRecognition:
module_name: text_recognition
model_name: PP-OCRv5_server_rec
model_dir: /app/server/paddleocr_v5/official_models/PP-OCRv5_server_rec
对的，想知道。文本检测模型 置信度怎么输出，
On Mon, Dec 1, 2025 at 11:03 AM Lin Manhui @.***> wrote:
Bobholamovic left a comment (PaddlePaddle/PaddleX#4775) https://github.com/PaddlePaddle/PaddleX/issues/4775#issuecomment-3594326809
请问你指的是高稳定性服务化部署方案吗？
— Reply to this email directly, view it on GitHub https://github.com/PaddlePaddle/PaddleX/issues/4775#issuecomment-3594326809, or unsubscribe https://github.com/notifications/unsubscribe-auth/BX6B7OPE4PCIJG6NBI7P3HD37OVXTAVCNFSM6AAAAACNNWFHHOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKOJUGMZDMOBQHE . You are receiving this because you authored the thread.Message ID: @.***>
比如卡证的识别，最后识别的置信度需要综合检测模型和识别模型的两个置信度的，单独看一个识别的置信度的话，如果哪个框标错了，也会影响整体识别的准确率。
On Mon, Dec 1, 2025 at 11:22 AM Lin Manhui @.***> wrote:
Bobholamovic left a comment (PaddlePaddle/PaddleX#4775) https://github.com/PaddlePaddle/PaddleX/issues/4775#issuecomment-3594361637
目前确实不支持输出文本检测的置信度；服务返回结果中只包含文本识别的置信度。请问需要文本检测的置信度用作什么处理？
— Reply to this email directly, view it on GitHub https://github.com/PaddlePaddle/PaddleX/issues/4775#issuecomment-3594361637, or unsubscribe https://github.com/notifications/unsubscribe-auth/BX6B7OPI6LCCCJK6ONGMVWT37OX53AVCNFSM6AAAAACNNWFHHOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKOJUGM3DCNRTG4 . You are receiving this because you authored the thread.Message ID: @.***>
目前我们支持通过textDetBoxThresh参数设置检测框置信度阈值，可用于筛掉低置信度的检测框。可以看看调整这个参数对你的场景是否有帮助。