Skip to content

Latest commit

 

History

History
172 lines (116 loc) · 5.07 KB

README_en.md

File metadata and controls

172 lines (116 loc) · 5.07 KB

faster_rcnn_resnet50_coco2017

Module Name faster_rcnn_resnet50_coco2017
Category object detection
Network faster_rcnn
Dataset COCO2017
Fine-tuning supported or not No
Module Size 131MB
Latest update date 2021-03-15
Data indicators -

I.Basic Information

  • Application Effect Display

    • Sample results：


  • Module Introduction

    • Faster_RCNN is a two-stage detector, it consists of feature extraction, proposal, classification and refinement processes. This module is trained on COCO2017 dataset, and can be used for object detection.

II.Installation

III.Module API Prediction

  • 1、Command line Prediction

    • $ hub run faster_rcnn_resnet50_coco2017 --input_path "/PATH/TO/IMAGE"
    • If you want to call the Hub module through the command line, please refer to: PaddleHub Command Line Instruction

  • 2、Prediction Code Example

    • import paddlehub as hub
import cv2

object_detector = hub.Module(name="faster_rcnn_resnet50_coco2017")
result = object_detector.object_detection(images=[cv2.imread('/PATH/TO/IMAGE')])
# or
# result = object_detector.object_detection(paths=['/PATH/TO/IMAGE'])

  • 3、API

    • def object_detection(paths=None,
                     images=None,
                     batch_size=1,
                     use_gpu=False,
                     output_dir='detection_result',
                     score_thresh=0.5,
                     visualization=True)

      • Detection API, detect positions of all objects in image

      • Parameters

        • paths (list[str]): image path;
        • images (list[numpy.ndarray]): image data, ndarray.shape is in the format [H, W, C], BGR;
        • batch_size (int): the size of batch;
        • use_gpu (bool): use GPU or not; set the CUDA_VISIBLE_DEVICES environment variable first if you are using GPU
        • output_dir (str): save path of images;
        • score_thresh (float): confidence threshold；
        • visualization (bool): Whether to save the results as picture files;

        NOTE: choose one parameter to provide data from paths and images

      • Return

        • res (list[dict]): results
          • data (list): detection results, each element in the list is dict
            • confidence (float): the confidence of the result
            • label (str): label
            • left (int): the upper left corner x coordinate of the detection box
            • top (int): the upper left corner y coordinate of the detection box
            • right (int): the lower right corner x coordinate of the detection box
            • bottom (int): the lower right corner y coordinate of the detection box
          • save_path (str, optional): output path for saving results
    • def save_inference_model(dirname)

      • Save model to specific path

      • Parameters

        • dirname: model save path

IV.Server Deployment

  • PaddleHub Serving can deploy an online service of object detection.

  • Step 1: Start PaddleHub Serving

    • Run the startup command：

    • $ hub serving start -m faster_rcnn_resnet50_coco2017

    • The servitization API is now deployed and the default port number is 8866.

    • NOTE: If GPU is used for prediction, set CUDA_VISIBLE_DEVICES environment variable before the service, otherwise it need not be set.

  • Step 2: Send a predictive request

    • With a configured server, use the following lines of code to send the prediction request and obtain the result

    • import requests
import json
import cv2
import base64


def cv2_to_base64(image):
  data = cv2.imencode('.jpg', image)[1]
  return base64.b64encode(data.tostring()).decode('utf8')

# Send an HTTP request
data = {'images':[cv2_to_base64(cv2.imread("/PATH/TO/IMAGE"))]}
headers = {"Content-type": "application/json"}
url = "http://127.0.0.1:8866/predict/faster_rcnn_resnet50_coco2017"
r = requests.post(url=url, headers=headers, data=json.dumps(data))

# print prediction results
print(r.json()["results"])

V.Release Note

  • 1.1.0

    First release

  • 1.1.1

    Fix the problem of reading numpy

  • 1.2.0

    Remove fluid api

    • $ hub install faster_rcnn_resnet50_coco2017==1.2.0