百度飛槳（5）—— 人臉識別切換電腦窗口顯示

前言

本節使用百度飛漿的預訓練模型RESNET-50訓練分類問題，這裏實現的是識別科比打開pycahrm識別到庫裏打開chrome。

效果圖

運行環境

ubuntu16.04
pycharm2019
paddlepaddle1.8.1

命令行切換窗口

首先安裝工具wmctrl 在linux的終端下執行：

sudo apt-get install wmctrl

切換窗口的命令：

wmctrl -a "pycharm"

主要任務

• 任務一：使用opencv讀取本地攝像頭
• 任務二：讀取攝像頭數據幀放入網絡識別匹配識別結果
• 任務三：根據識別結果切換窗口

這裏因爲opencv讀取攝像頭數據比較佔用資源，數據幀識別也比較佔資源，所以我開啓了兩個線程處理這兩個問題，定義了一個標誌位，只有數據幀處理完後才寫入本地攝像頭數據，寫入後才處理數據幀。

訓練網絡代碼

這個就是配置網絡的一些參數，導入網絡模型，基本是固定格式。 train.py文件如下：

import paddlehub as hub
from paddlehub.dataset.base_cv_dataset import BaseCVDataset

class DemoDataset(BaseCVDataset):
    def __init__(self):
        # 數據集存放位置

        self.dataset_dir = ""
        super(DemoDataset, self).__init__(
            base_path=self.dataset_dir,
            train_list_file="dataset/train_list.txt",
            validate_list_file="dataset/validate_list.txt",
            test_list_file="dataset/test_list.txt",
            label_list_file="dataset/label_list.txt",
        )

module = hub.Module(name="resnet_v2_50_imagenet")
dataset = DemoDataset()

data_reader = hub.reader.ImageClassificationReader(
    image_width=module.get_expected_image_width(),
    image_height=module.get_expected_image_height(),
    images_mean=module.get_pretrained_images_mean(),
    images_std=module.get_pretrained_images_std(),
    dataset=dataset)

config = hub.RunConfig(
    use_cuda=False,  # 是否使用GPU訓練，默認爲False；
    num_epoch=5,  # Fine-tune的輪數；
    checkpoint_dir="cv_finetune_turtorial_demo",  # 模型checkpoint保存路徑, 若用戶沒有指定，程序會自動生成；
    batch_size=10,  # 訓練的批大小，如果使用GPU，請根據實際情況調整batch_size；
    eval_interval=10,  # 模型評估的間隔，默認每100個step評估一次驗證集；
    strategy=hub.finetune.strategy.DefaultFinetuneStrategy())  #Fine-tune優化策略；
    #strategy=hub.finetune.strategy.AdamWeightDecayStrategy())

input_dict, output_dict, program = module.context(trainable=True)
img = input_dict["image"]
feature_map = output_dict["feature_map"]
feed_list = [img.name]

task = hub.ImageClassifierTask(
    data_reader=data_reader,
    feed_list=feed_list,
    feature=feature_map,
    num_classes=dataset.num_labels,
    config=config)

run_states = task.finetune_and_eval()

任務一代碼

顯示攝像頭數據並且寫入本地

def showImg():
    global flag
    cap = cv2.VideoCapture(0)
    while 1:
        ret,frame = cap.read()
        cv2.imshow("cap",frame)

        if flag is 0:
            cv2.imwrite("./temp_out/cap.jpg",frame)
            flag = 1

        if cv2.waitKey(100) & 0xff == ord('q'):
            break
    cap.release()
    cv2.destroyAllWindows()

任務二代碼

識別數據幀並切換窗口

def recognize():
    global flag
    module = hub.Module(name="resnet_v2_50_imagenet")
    dataset = DemoDataset()

    data_reader = hub.reader.ImageClassificationReader(
        image_width=module.get_expected_image_width(),
        image_height=module.get_expected_image_height(),
        images_mean=module.get_pretrained_images_mean(),
        images_std=module.get_pretrained_images_std(),
        dataset=dataset)

    config = hub.RunConfig(
        use_cuda=False,  # 是否使用GPU訓練，默認爲False；
        num_epoch=5,  # Fine-tune的輪數；
        checkpoint_dir="cv_finetune_turtorial_demo",  # 模型checkpoint保存路徑, 若用戶沒有指定，程序會自動生成；
        batch_size=10,  # 訓練的批大小，如果使用GPU，請根據實際情況調整batch_size；
        eval_interval=10,  # 模型評估的間隔，默認每100個step評估一次驗證集；
        strategy=hub.finetune.strategy.DefaultFinetuneStrategy())  #Fine-tune優化策略；
        #strategy=hub.finetune.strategy.AdamWeightDecayStrategy())

    input_dict, output_dict, program = module.context(trainable=True)
    img = input_dict["image"]
    feature_map = output_dict["feature_map"]
    feed_list = [img.name]

    task = hub.ImageClassifierTask(
        data_reader=data_reader,
        feed_list=feed_list,
        feature=feature_map,
        num_classes=dataset.num_labels,
        config=config)

    label_map = dataset.label_dict()
    #run_states = task.finetune_and_eval()
    while 1:
        if flag is 1:
            data = []
            data.append("/home/xmy/PycharmProjects/test/paddle/proj3_recognizeMyself/temp_out/cap.jpg")
            index = 0
            run_states = task.predict(data=data)
            results = [run_state.run_results for run_state in run_states]

            for batch_result in results:
                batch_result = np.argmax(batch_result, axis=2)[0]
                for result in batch_result:
                    index += 1
                    result = label_map[result]
                    #print("input %i is %s, and the predict result is %s" %
                        #(index, data[index - 1], result))

            if "科比" in result:
                os.system("wmctrl -a \"pycharm\"")
            elif "庫裏" in result:
                os.system("wmctrl -a \"chrome\"")
            flag = 0

完整識別代碼

import paddlehub as hub
from paddlehub.dataset.base_cv_dataset import BaseCVDataset
import numpy as np
import cv2
import threading
import os

flag = 0

class DemoDataset(BaseCVDataset):
    def __init__(self):
        # 數據集存放位置

        self.dataset_dir = ""
        super(DemoDataset, self).__init__(
            base_path=self.dataset_dir,
            train_list_file="dataset/train_list.txt",
            validate_list_file="dataset/validate_list.txt",
            test_list_file="dataset/test_list.txt",
            label_list_file="dataset/label_list.txt",
        )


def showImg():
    global flag
    cap = cv2.VideoCapture(0)
    while 1:
        ret,frame = cap.read()
        cv2.imshow("cap",frame)

        if flag is 0:
            cv2.imwrite("./temp_out/cap.jpg",frame)
            flag = 1

        if cv2.waitKey(100) & 0xff == ord('q'):
            break
    cap.release()
    cv2.destroyAllWindows()

def recognize():
    global flag
    module = hub.Module(name="resnet_v2_50_imagenet")
    dataset = DemoDataset()

    data_reader = hub.reader.ImageClassificationReader(
        image_width=module.get_expected_image_width(),
        image_height=module.get_expected_image_height(),
        images_mean=module.get_pretrained_images_mean(),
        images_std=module.get_pretrained_images_std(),
        dataset=dataset)

    config = hub.RunConfig(
        use_cuda=False,  # 是否使用GPU訓練，默認爲False；
        num_epoch=5,  # Fine-tune的輪數；
        checkpoint_dir="cv_finetune_turtorial_demo",  # 模型checkpoint保存路徑, 若用戶沒有指定，程序會自動生成；
        batch_size=10,  # 訓練的批大小，如果使用GPU，請根據實際情況調整batch_size；
        eval_interval=10,  # 模型評估的間隔，默認每100個step評估一次驗證集；
        strategy=hub.finetune.strategy.DefaultFinetuneStrategy())  #Fine-tune優化策略；
        #strategy=hub.finetune.strategy.AdamWeightDecayStrategy())

    input_dict, output_dict, program = module.context(trainable=True)
    img = input_dict["image"]
    feature_map = output_dict["feature_map"]
    feed_list = [img.name]

    task = hub.ImageClassifierTask(
        data_reader=data_reader,
        feed_list=feed_list,
        feature=feature_map,
        num_classes=dataset.num_labels,
        config=config)

    label_map = dataset.label_dict()
    #run_states = task.finetune_and_eval()
    while 1:
        if flag is 1:
            data = []
            data.append("/home/xmy/PycharmProjects/test/paddle/proj3_recognizeMyself/temp_out/cap.jpg")
            index = 0
            run_states = task.predict(data=data)
            results = [run_state.run_results for run_state in run_states]

            for batch_result in results:
                batch_result = np.argmax(batch_result, axis=2)[0]
                for result in batch_result:
                    index += 1
                    result = label_map[result]
                    #print("input %i is %s, and the predict result is %s" %
                        #(index, data[index - 1], result))

            if "科比" in result:
                os.system("wmctrl -a \"pycharm\"")
            elif "庫裏" in result:
                os.system("wmctrl -a \"chrome\"")
            flag = 0



if __name__ == '__main__':
    t1 = threading.Thread(target=showImg)
    t2 = threading.Thread(target=recognize)
    t1.start()
    t2.start()

完整項目以及數據集以及訓練好的模型下載

github地址
使用代碼記得修改代碼裏的路徑因爲使用相對路徑paddle總是報錯找不到… …我是真鬱悶所以所有路徑都是寫的絕對路徑。