paddleocr识别结果比rapidocr效果较好，看看是什么原因

[scrtou]

识别图片如下，希望识别出右下角数字7

rapidcor识别结果

参数：

Global:
    text_score: 0

    use_det: true
    use_cls: false
    use_rec: true

    min_height: 5
    width_height_ratio: -1
    max_side_len: 2000
    min_side_len: 5

    return_word_box: true
    return_single_char_box: true

    font_path: null
    log_level: "info" # debug / info / warning / error / critical

EngineConfig:
    onnxruntime:
        intra_op_num_threads: -1
        inter_op_num_threads: -1
        enable_cpu_mem_arena: false

        cpu_ep_cfg:
            arena_extend_strategy: "kSameAsRequested"

        use_cuda: false
        cuda_ep_cfg:
            device_id: 0
            arena_extend_strategy: "kNextPowerOfTwo"
            cudnn_conv_algo_search: "EXHAUSTIVE"
            do_copy_in_default_stream: true

        use_dml: false
        dm_ep_cfg: null

        use_cann: false
        cann_ep_cfg:
            device_id: 0
            arena_extend_strategy: "kNextPowerOfTwo"
            npu_mem_limit:  21474836480 # 20 * 1024 * 1024 * 1024
            op_select_impl_mode: "high_performance"
            optypelist_for_implmode: "Gelu"
            enable_cann_graph: true

    openvino:
        inference_num_threads: -1
        performance_hint: null
        performance_num_requests: -1
        enable_cpu_pinning: null
        num_streams: -1
        enable_hyper_threading: null
        scheduling_core_type: null

    paddle:
        cpu_math_library_num_threads: -1
        use_npu: false
        npu_id: 0
        use_cuda: false
        gpu_id: 0
        gpu_mem: 500

    torch:
        use_cuda: false
        gpu_id: 0

Det:
    engine_type: "onnxruntime"
    lang_type: "ch"
    model_type: "mobile"
    ocr_version: "PP-OCRv5"

    task_type: "det"

    model_path: null
    model_dir: null

    limit_side_len: 1280
    limit_type: min
    std: [ 0.5, 0.5, 0.5 ]
    mean: [ 0.5, 0.5, 0.5 ]

    thresh: 0.7
    box_thresh: 0.1
    max_candidates: 2000
    unclip_ratio: 2
    use_dilation: true
    score_mode: slow

Cls:
    engine_type: "onnxruntime"
    lang_type: "ch"
    model_type: "mobile"
    ocr_version: "PP-OCRv4"

    task_type: "cls"

    model_path: null
    model_dir: null

    cls_image_shape: [3, 48, 192]
    cls_batch_num: 6
    cls_thresh: 0.5
    label_list: ["0", "180"]

Rec:
    engine_type: "onnxruntime"
    lang_type: "ch"
    model_type: "mobile"
    ocr_version: "PP-OCRv5"

    task_type: "rec"

    model_path: null
    model_dir: null

    rec_keys_path: null
    rec_img_shape: [3, 48, 320]
    rec_batch_num: 6

paddleocr识别结果：

参数：

SubModules:
  TextDetection:
    box_thresh: 0.1
    limit_side_len: 1280
    limit_type: min
    max_side_limit: 2000
    model_dir: null
    model_name: PP-OCRv5_mobile_det
    module_name: text_detection
    thresh: 0.7
    unclip_ratio: 2
  TextLineOrientation:
    batch_size: 6
    model_dir: null
    model_name: PP-LCNet_x1_0_textline_ori
    module_name: textline_orientation
  TextRecognition:
    batch_size: 6
    model_dir: null
    model_name: PP-OCRv5_mobile_rec
    module_name: text_recognition
    score_thresh: 0.0
SubPipelines:
  DocPreprocessor:
    SubModules:
      DocOrientationClassify:
        model_dir: null
        model_name: PP-LCNet_x1_0_doc_ori
        module_name: doc_text_orientation
      DocUnwarping:
        model_dir: null
        model_name: UVDoc
        module_name: image_unwarping
    pipeline_name: doc_preprocessor
    use_doc_orientation_classify: false
    use_doc_unwarping: false
pipeline_name: OCR
text_type: general
use_doc_preprocessor: false
use_textline_orientation: false

[scrtou]

@SWHL

[SWHL]

这种轻微差别难以排查。需要给出具体详尽的环境，各个依赖包的版本来看。还有可能是模型转换精度问题。