FastDeploy项目简介，使用其进行（图像分类、目标检测、语义分割、文本检测|orc部署）

万里鹏程转瞬至2024-01-29 9:18

FastDeploy是一款全场景、易用灵活、极致高效的AI推理部署工具， 支持云边端部署。提供超过 🔥160+ Text，Vision， Speech和跨模态模型📦开箱即用的部署体验，并实现🔚端到端的推理性能优化。包括 物体检测、字符识别（OCR）、人脸、人像扣图、多目标跟踪系统、NLP、Stable Diffusion文图生成、TTS 等几十种任务场景，满足开发者多场景、多硬件、多平台的产业部署需求。

1、FastDeploy支持的环境

1.1 芯片支持

FastDeploy支持各个常见的硬件平台和各种国产化的芯片。例如华为的ascend
X86_64 CPU
NVDIA GPU
飞腾 CPU
昆仑芯 XPU
华为昇腾 NPU
Graphcore IPU
算能
Intel 显卡
Jetson
ARM CPU
RK3588等
RV1126等
晶晨
恩智浦

1.2 编程语言支持

FastDeploy支持python、c++、java等开发语言，同时还支持web与小程序部署。

1.3 操作系统支持

FastDeploy支持linux、windows、mac、安卓部署。

1.4 推理后端支持

FastDeploy对各种推理后端接口均做了封装，可以以一种代码风格同时调用onnx、paddle infer、tensorrt、openvino等推理框架。

2、安装与使用

本博文对应的代码和模型资源可以进行付费下载，也可以参考博文资料轻松实现复现。

2.1 安装命令

🔸 前置依赖
  • CUDA >= 11.2、cuDNN >= 8.0、Python >= 3.6
  • OS: Linux x86_64/macOS/Windows 10
🔸 安装GPU版本
bash 复制代码 
pip install numpy opencv-python fastdeploy-gpu-python -f https://www.paddlepaddle.org.cn/whl/fastdeploy.html
🔸 Conda安装(推荐✨)
bash 复制代码 
conda config --add channels conda-forge && conda install cudatoolkit=11.2 cudnn=8.2
🔸 安装CPU版本
bash 复制代码 
pip install numpy opencv-python fastdeploy-python -f https://www.paddlepaddle.org.cn/whl/fastdeploy.html

最终安装成功的输出如下所示

2.2 部署图像分类模型

下载paddle官方的预训练模型，https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz，也可以使用自己基于paddleclas训练的模型

将以下图片保存为000000014439.jpg

预测代码如下所示，使用onnxruntime-gpu进行推理

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy as fd
import fastdeploy.vision as vision
import ipdb

option = fd.RuntimeOption()
#option.use_trt_backend()
option.use_gpu(device_id=0) #使用0号gpu进行预测
option.use_ort_backend() #使用onnxruntime作为后端

model_file="ResNet50_vd_infer/inference.pdmodel"
params_file="ResNet50_vd_infer/inference.pdiparams"
config_file="ResNet50_vd_infer/inference_cls.yaml"
model = fd.vision.classification.PaddleClasModel(
    model_file, params_file, config_file, runtime_option=option)
im = cv2.imread("000000014439.jpg")
result = model.predict(im, topk=5)
print(type(result),dir(result))
print(result)

程序运行后输出如下所示：

py 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW                                                                  Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert                                                            BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
[INFO] fastdeploy/runtime/runtime.cc(300)::fastdeploy::Runtime::CreateOrtBackend                                                                                       Runtime initialized with Backend::ORT in Device::GPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.ClassifyResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', 'feature', 'label_ids', 'scores']
ClassifyResult(
label_ids: 701, 879, 417, 645, 723, 
scores: 0.424514, 0.084369, 0.058738, 0.016847, 0.013349, 
)

从以上输出中可以看到，预测结果又'feature', 'label_ids', 'scores'三个属性。

2.2 部署目标检测模型

下载paddle官方的预训练模型，也可以使用自己导出的paddle infer模型。
https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz

将以下图片保存为000000014439.jpg

预测代码如下所示

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy.vision as vision

model = vision.detection.PPYOLOE("ppyoloe_crn_l_300e_coco/model.pdmodel",
                                 "ppyoloe_crn_l_300e_coco/model.pdiparams",
                                 "ppyoloe_crn_l_300e_coco/infer_cfg.yml")
im = cv2.imread("000000014439.jpg")
result = model.predict(im)
print(result)

vis_im = vision.vis_detection(im, result, score_threshold=0.5)
cv2.imshow("vis_image.jpg", vis_im)
cv2.waitKey()
#cv2.imwrite("vis_image.jpg", vis_im)

程序运行时的输出如下所示，可见默认是使用OPENVINO cpu进行推理。其返回结果为DetectionResult对象，其一共有'boxes', 'contain_masks', 'label_ids', 'masks', 'rotated_boxes', 'scores'等属性。其中boxes是xmin, ymin, xmax, ymax,格式的数据，其默认只返回nms操作后的top 300个结果，后续需要自己根据scores进行结果过滤。

txt 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(45)::fastdeploy::vision::FuseNormalizeCast      Normalize and Cast are fused to Normalize in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW   Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert     BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle     number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle     affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle     Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend   Runtime initialized with Backend::OPENVINO in Device::CPU.
DetectionResult: [xmin, ymin, xmax, ymax, score, label_id]
415.047180,89.311569, 506.009613, 283.863098, 0.950423, 0
163.665710,81.914932, 198.585342, 166.760895, 0.896433, 0
581.788635,113.027618, 612.623474, 198.521713, 0.842596, 0
267.217224,89.777306, 298.796051, 169.361526, 0.837951, 0
104.465584,45.482422, 127.688850, 93.533867, 0.773348, 0
348.902130,44.059338, 367.541687, 98.403542, 0.767127, 0
363.889740,58.385532, 381.397522, 114.652367, 0.756467, 0
504.843811,114.531601, 612.280945, 271.292572, 0.714129, 0

2.4 部署语义分割模型

下载paddle发布的预训练模型SegFormer_B0，也可以使用paddleseg训练后导出模型。

https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz

将以下图片保存为000000014439.jpg

预测代码如下所示

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy as fd
import fastdeploy.vision as vision
import numpy as np
import ipdb
def get_color_map_list(num_classes):
    """
    Args:
        num_classes (int): number of class
    Returns:
        color_map (list): RGB color list
    """
    color_map = num_classes * [0, 0, 0]
    for i in range(0, num_classes):
        j = 0
        lab = i
        while lab:
            color_map[i * 3] |= (((lab >> 0) & 1) << (7 - j))
            color_map[i * 3 + 1] |= (((lab >> 1) & 1) << (7 - j))
            color_map[i * 3 + 2] |= (((lab >> 2) & 1) << (7 - j))
            j += 1
            lab >>= 3
    color_map = [color_map[i:i + 3] for i in range(0, len(color_map), 3)]
    return color_map

color_map=get_color_map_list(80) #80个类别
color_map=np.array(color_map).astype(np.uint8)

option = fd.RuntimeOption()
#option.use_kunlunxin() #使用昆仑芯片进行部署
#option.use_lite_backend()
option.use_paddle_infer_backend()

model_file="SegFormer_B0-cityscapes-with-argmax/model.pdmodel"
params_file="SegFormer_B0-cityscapes-with-argmax/model.pdiparams"
config_file="SegFormer_B0-cityscapes-with-argmax/deploy.yaml"
model = fd.vision.segmentation.PaddleSegModel(
    model_file, params_file, config_file, runtime_option=option)
im = cv2.imread("000000014439.jpg")
result = model.predict(im)
print(type(result),dir(result))
print(result)
#ipdb.set_trace()
res=np.array(result.label_map,np.uint8)
res=res.reshape(result.shape)

res=color_map[res] #为结果上色
cv2.imshow("label_map",res)
cv2.waitKey()

代码运行时输出如下所示：

py 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW                                                                  Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert                                                            BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0125 21:32:56.195503  5356 analysis_config.cc:971] It is detected that mkldnn and memory_optimize_pass are enabled at the same time, but they are not supported yet. Currently, memory_optimize_pass is explicitly disabled
[INFO] fastdeploy/runtime/runtime.cc(273)::fastdeploy::Runtime::CreatePaddleBackend                                                                                    Runtime initialized with Backend::PDINFER in Device::CPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.SegmentationResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', 'contain_score_map', 'label_map', 'score_map', 'shape']
SegmentationResult Image masks 10 rows x 10 cols:
[4, 4, 4, 4, 4, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 4, 4, 4, .....]
[4, 4, 4, 4, 4, 4, 4, 4, 4, 4, .....]
[4, 4, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
...........
result shape is: [404 640]

运行效果如下所示

2.5 部署文本识别模型

进行全流程文本识别

首先下载以下模型，可以手动打开http链接进行下载。

下载PP-OCRv3文字检测模型

wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_infer.tar
tar -xvf ch_PP-OCRv3_det_infer.tar

下载文字方向分类器模型

wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar
tar -xvf ch_ppocr_mobile_v2.0_cls_infer.tar

下载PP-OCRv3文字识别模型

wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_rec_infer.tar
tar -xvf ch_PP-OCRv3_rec_infer.tar

准备label文件

下载ppocr_keys_v1.txt，保存到代码目录下。

准备测试文件

将以下图片保存为12.jpg

测试代码如下所示;

py 复制代码 
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import fastdeploy as fd
import cv2
import os

def parse_arguments():
    import argparse
    import ast
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--det_model", default="ch_PP-OCRv3_det_infer", help="Path of Detection model of PPOCR.")
    parser.add_argument(
        "--cls_model",default="ch_ppocr_mobile_v2.0_cls_infer",
        help="Path of Classification model of PPOCR.")
    parser.add_argument(
        "--rec_model",default="ch_PP-OCRv3_rec_infer",
        help="Path of Recognization model of PPOCR.")
    parser.add_argument(
        "--rec_label_file",
        default="ppocr_keys_v1.txt",
        help="Path of Recognization model of PPOCR.")
    parser.add_argument(
        "--image", type=str, default="12.jpg", help="Path of test image file.")
    parser.add_argument(
        "--device",
        type=str,
        default='cpu',
        help="Type of inference device, support 'cpu' or 'gpu'.")
    parser.add_argument(
        "--device_id",
        type=int,
        default=0,
        help="Define which GPU card used to run model.")
    parser.add_argument(
        "--cls_bs",
        type=int,
        default=1,
        help="Classification model inference batch size.")
    parser.add_argument(
        "--rec_bs",
        type=int,
        default=6,
        help="Recognition model inference batch size")
    parser.add_argument(
        "--backend",
        type=str,
        default="openvino",
        help="Type of inference backend, support ort/trt/paddle/openvino, default 'openvino' for cpu, 'tensorrt' for gpu"
    )

    return parser.parse_args()


def build_option(args):

    det_option = fd.RuntimeOption()
    cls_option = fd.RuntimeOption()
    rec_option = fd.RuntimeOption()

    if args.device.lower() == "gpu":
        det_option.use_gpu(args.device_id)
        cls_option.use_gpu(args.device_id)
        rec_option.use_gpu(args.device_id)

    if args.backend.lower() == "trt":
        assert args.device.lower(
        ) == "gpu", "TensorRT backend require inference on device GPU."
        det_option.use_trt_backend()
        cls_option.use_trt_backend()
        rec_option.use_trt_backend()

        # If use TRT backend, the dynamic shape will be set as follow.
        # We recommend that users set the length and height of the detection model to a multiple of 32.
        # We also recommend that users set the Trt input shape as follow.
        det_option.set_trt_input_shape("x", [1, 3, 64, 64], [1, 3, 640, 640],
                                       [1, 3, 960, 960])
        cls_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.cls_bs, 3, 48, 320],
                                       [args.cls_bs, 3, 48, 1024])
        rec_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.rec_bs, 3, 48, 320],
                                       [args.rec_bs, 3, 48, 2304])

        # Users could save TRT cache file to disk as follow.
        det_option.set_trt_cache_file(args.det_model + "/det_trt_cache.trt")
        cls_option.set_trt_cache_file(args.cls_model + "/cls_trt_cache.trt")
        rec_option.set_trt_cache_file(args.rec_model + "/rec_trt_cache.trt")

    elif args.backend.lower() == "pptrt":
        assert args.device.lower(
        ) == "gpu", "Paddle-TensorRT backend require inference on device GPU."
        det_option.use_paddle_infer_backend()
        det_option.paddle_infer_option.collect_trt_shape = True
        det_option.paddle_infer_option.enable_trt = True

        cls_option.use_paddle_infer_backend()
        cls_option.paddle_infer_option.collect_trt_shape = True
        cls_option.paddle_infer_option.enable_trt = True

        rec_option.use_paddle_infer_backend()
        rec_option.paddle_infer_option.collect_trt_shape = True
        rec_option.paddle_infer_option.enable_trt = True

        # If use TRT backend, the dynamic shape will be set as follow.
        # We recommend that users set the length and height of the detection model to a multiple of 32.
        # We also recommend that users set the Trt input shape as follow.
        det_option.set_trt_input_shape("x", [1, 3, 64, 64], [1, 3, 640, 640],
                                       [1, 3, 960, 960])
        cls_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.cls_bs, 3, 48, 320],
                                       [args.cls_bs, 3, 48, 1024])
        rec_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.rec_bs, 3, 48, 320],
                                       [args.rec_bs, 3, 48, 2304])

        # Users could save TRT cache file to disk as follow.
        det_option.set_trt_cache_file(args.det_model)
        cls_option.set_trt_cache_file(args.cls_model)
        rec_option.set_trt_cache_file(args.rec_model)

    elif args.backend.lower() == "ort":
        det_option.use_ort_backend()
        cls_option.use_ort_backend()
        rec_option.use_ort_backend()

    elif args.backend.lower() == "paddle":
        det_option.use_paddle_infer_backend()
        cls_option.use_paddle_infer_backend()
        rec_option.use_paddle_infer_backend()

    elif args.backend.lower() == "openvino":
        assert args.device.lower(
        ) == "cpu", "OpenVINO backend require inference on device CPU."
        det_option.use_openvino_backend()
        cls_option.use_openvino_backend()
        rec_option.use_openvino_backend()

    elif args.backend.lower() == "pplite":
        assert args.device.lower(
        ) == "cpu", "Paddle Lite backend require inference on device CPU."
        det_option.use_lite_backend()
        cls_option.use_lite_backend()
        rec_option.use_lite_backend()

    return det_option, cls_option, rec_option


args = parse_arguments()


det_model_file = os.path.join(args.det_model, "inference.pdmodel")
det_params_file = os.path.join(args.det_model, "inference.pdiparams")

cls_model_file = os.path.join(args.cls_model, "inference.pdmodel")
cls_params_file = os.path.join(args.cls_model, "inference.pdiparams")

rec_model_file = os.path.join(args.rec_model, "inference.pdmodel")
rec_params_file = os.path.join(args.rec_model, "inference.pdiparams")
rec_label_file = args.rec_label_file

det_option, cls_option, rec_option = build_option(args)

det_model = fd.vision.ocr.DBDetector(
    det_model_file, det_params_file, runtime_option=det_option)

cls_model = fd.vision.ocr.Classifier(
    cls_model_file, cls_params_file, runtime_option=cls_option)

rec_model = fd.vision.ocr.Recognizer(
    rec_model_file, rec_params_file, rec_label_file, runtime_option=rec_option)

# Parameters settings for pre and post processing of Det/Cls/Rec Models.
# All parameters are set to default values.
det_model.preprocessor.max_side_len = 960
det_model.postprocessor.det_db_thresh = 0.3
det_model.postprocessor.det_db_box_thresh = 0.6
det_model.postprocessor.det_db_unclip_ratio = 1.5
det_model.postprocessor.det_db_score_mode = "slow"
det_model.postprocessor.use_dilation = False
cls_model.postprocessor.cls_thresh = 0.9

# Create PP-OCRv3, if cls_model is not needed, just set cls_model=None .
ppocr_v3 = fd.vision.ocr.PPOCRv3(
    det_model=det_model, cls_model=cls_model, rec_model=rec_model)

# Set inference batch size for cls model and rec model, the value could be -1 and 1 to positive infinity.
# When inference batch size is set to -1, it means that the inference batch size
# of the cls and rec models will be the same as the number of boxes detected by the det model.
ppocr_v3.cls_batch_size = args.cls_bs
ppocr_v3.rec_batch_size = args.rec_bs

# Read the input image
im = cv2.imread(args.image)

# Predict and reutrn the results
result = ppocr_v3.predict(im)

print(result)

# Visuliaze the results.
vis_im = fd.vision.vis_ppocr(im, result)
cv2.imwrite("visualized_result.jpg", vis_im)
cv2.imshow("visualized_result.jpg", vis_im)
cv2.waitKey()
print("Visualized result save in ./visualized_result.jpg")

以上代码基于paddleocr的官方代码改进而来，故而可以参考官方的命令参数进行使用

```# 在CPU上使用Paddle Inference推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend paddle
# 在CPU上使用OenVINO推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend openvino
# 在CPU上使用ONNX Runtime推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend ort
# 在CPU上使用Paddle Lite推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend pplite
# 在GPU上使用Paddle Inference推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend paddle
# 在GPU上使用Paddle TensorRT推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend pptrt
# 在GPU上使用ONNX Runtime推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend ort
# 在GPU上使用Nvidia TensorRT推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend trt

代码运行效果图如下所示

程序运行时输出如下所示，可以看到大部分文字都准确识别了。

py 复制代码 
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
det boxes: [[42,413],[483,391],[484,428],[43,450]]rec text: 上海斯格威铂尔大酒店 rec score:0.989301 cls label: 0 cls score: 1.000000
det boxes: [[187,456],[399,448],[400,480],[188,488]]rec text: 打浦路15号 rec score:0.994266 cls label: 0 cls score: 1.000000
det boxes: [[23,507],[513,488],[515,529],[24,548]]rec text: 绿洲仕格维花园公寓 rec score:0.984679 cls label: 0 cls score: 1.000000
det boxes: [[74,553],[427,542],[428,571],[75,582]]rec text: 打浦路252935号 rec score:0.996439 cls label: 0 cls score: 1.000000
Visualized result save in ./visualized_result.jpg
对文本切片进行识别
模型简介 模型名称 推荐场景 检测模型 方向分类器 识别模型
中英文超轻量PP-OCRv4模型（15.8M） ch_PP-OCRv4_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型
中英文超轻量PP-OCRv3模型（16.2M） ch_PP-OCRv3_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型
英文超轻量PP-OCRv3模型（13.4M） en_PP-OCRv3_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型

各位可以自行截图生成如下图片，保存为13.jpg

识别代码如下所示，识别效果十分棒，各位可以换成自己的文本切片。

py 复制代码 
import fastdeploy as fd
import cv2
import os

rec_model_file = os.path.join('ch_ppocr_mobile_v2.0_cls_infer', "inference.pdmodel")
rec_params_file = os.path.join('ch_ppocr_mobile_v2.0_cls_infer', "inference.pdiparams")
rec_label_file = "ppocr_keys_v1.txt"

# Set the runtime option
rec_option =  fd.RuntimeOption()
rec_option.use_ort_backend()

# Create the rec_model
rec_model = fd.vision.ocr.Recognizer(
    rec_model_file, rec_params_file, rec_label_file, runtime_option=rec_option)

# Read the image
im = cv2.imread("13.jpg")

# Predict and return the result
result = rec_model.predict(im)
print(type(result),dir(result))
# User can infer a batch of images by following code.
# result = rec_model.batch_predict([im])

print(result)

识别结果如下所示，可以看到识别结果完全正确。

py 复制代码 
[INFO] fastdeploy/runtime/runtime.cc(300)::fastdeploy::Runtime::CreateOrtBackend        Runtime initialized with Backend::ORT in Device::CPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.OCRResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__str__', '__subclasshook__', 'boxes', 'cls_labels', 'cls_scores', 'rec_scores', 'text']
rec text: 绿洲仕格维花园公寓 rec score:0.988077

3、其他功能

前文仅描述了常见功能的python使用，在faskdeploy还提供了c++部署案列，其python接口与c++接口基本一致。

除了图像分类、目标检测、语义分割、文本识别等功能外，faskdeploy还提供了人脸检测、人脸特征点、人脸识别、生成式模型、姿态检测、扣图、图像超分、目标跟踪等功能的部署案例。后续有使用到会在这里补充案例。

相关推荐
吕小明么2 分钟前
OpenAI o3 “震撼” 发布后回归技术本身的审视与进一步思考
人工智能·深度学习·算法·aigc·agi
CSBLOG1 小时前
深度学习试题及答案解析（一）
人工智能·深度学习
小陈phd1 小时前
深度学习之超分辨率算法——SRCNN
python·深度学习·tensorflow·卷积
王国强20092 小时前
动手学人工智能-深度学习计算5-文件读写操作
深度学习
威化饼的一隅3 小时前
【多模态】swift-3框架使用
人工智能·深度学习·大模型·swift·多模态
机器学习之心4 小时前
BiTCN-BiGRU基于双向时间卷积网络结合双向门控循环单元的数据多特征分类预测（多输入单输出）
深度学习·分类·gru
发呆小天才O.oᯅ4 小时前
YOLOv8目标检测——详细记录使用OpenCV的DNN模块进行推理部署C++实现
c++·图像处理·人工智能·opencv·yolo·目标检测·dnn
深度学习lover5 小时前
＜项目代码＞YOLO Visdrone航拍目标识别＜目标检测＞
python·yolo·目标检测·计算机视觉·visdrone航拍目标识别
MorleyOlsen5 小时前
【Trick】解决服务器cuda报错——RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED
运维·服务器·深度学习
愚者大大7 小时前
1. 深度学习介绍
人工智能·深度学习
热门推荐
01〔AI 绘画〕Stable Diffusion 之 解决绘制多人或面部很小的人物时面部崩坏问题 篇02YOLOv8训练好模型后，追加轮数继续训练、或者提前终止训练，缩减训练轮数03玄机平台应急响应—webshell查杀04机器人编程怎么入门？05HCIA-datacom数通题库和录播视频资料06Docker 夺命连环 15 问07SQL 主从数据库实时备份08校验 GPT-4 真实性的三个经典问题：快速区分 GPT-3.5 与 GPT-4，并提供免费测试网站09(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明10VSCode插件 —— Cody AI （免费AI助手！）