{"id":38915,"date":"2024-11-02T01:40:00","date_gmt":"2024-11-02T01:40:00","guid":{"rendered":"http:\/\/atmokpo.com\/w\/?p=38915"},"modified":"2024-11-26T06:42:58","modified_gmt":"2024-11-26T06:42:58","slug":"yolo%eb%a5%bc-%ec%9d%b4%ec%9a%a9%ed%95%9c-%ec%8b%a4%ec%8b%9c%ea%b0%84-%ea%b0%9d%ec%b2%b4-%ea%b2%80%ec%b6%9c-%ec%95%a0%ed%94%8c%eb%a6%ac%ec%bc%80%ec%9d%b4%ec%85%98-%ea%b5%ac%ec%b6%95-yolo%eb%a5%bc-ope","status":"publish","type":"post","link":"https:\/\/atmokpo.com\/w\/38915\/","title":{"rendered":"YOLO\ub97c \uc774\uc6a9\ud55c \uc2e4\uc2dc\uac04 \uac1d\uccb4 \uac80\ucd9c \uc560\ud50c\ub9ac\ucf00\uc774\uc158 \uad6c\ucd95, YOLO\ub97c OpenCV\uc640 \uc5f0\ub3d9\ud558\uc5ec \uc2e4\uc2dc\uac04 \uac1d\uccb4 \uac80\ucd9c \uc2dc\uc2a4\ud15c \ub9cc\ub4e4\uae30"},"content":{"rendered":"

<\/p>\n

\n \ubcf8 \uae00\uc5d0\uc11c\ub294 YOLO(You Only Look Once) \uc54c\uace0\ub9ac\uc998\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc2e4\uc2dc\uac04 \uac1d\uccb4 \uac80\ucd9c \uc560\ud50c\ub9ac\ucf00\uc774\uc158\uc744 \uad6c\ucd95\ud558\ub294 \ubc29\ubc95\uc744 \uc124\uba85\ud569\ub2c8\ub2e4.
\n YOLO\ub294 \uace0\uc18d\uc73c\ub85c \uac1d\uccb4\ub97c \ud0d0\uc9c0\ud560 \uc218 \uc788\ub294 \uc54c\uace0\ub9ac\uc998\uc73c\ub85c, \ub2e4\uc591\ud55c \ubd84\uc57c\uc5d0\uc11c \ud65c\uc6a9\ub418\uace0 \uc788\uc2b5\ub2c8\ub2e4.
\n OpenCV\uc640 \ud568\uaed8 YOLO\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc6f9\ucea0\uc5d0\uc11c \uc2e4\uc2dc\uac04\uc73c\ub85c \uac1d\uccb4\ub97c \uac80\ucd9c\ud558\ub294 \uc2dc\uc2a4\ud15c\uc744 \ub9cc\ub4dc\ub294 \uacfc\uc815\uc744 \uc790\uc138\ud788 \uc124\uba85\ud558\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n

1. YOLO\ub780?<\/h2>\n

\n YOLO\ub294 \uac1d\uccb4 \uac80\ucd9c\uc744 \uc704\ud574 \uc124\uacc4\ub41c \uc2e0\uacbd\ub9dd \uae30\ubc18\uc758 \uc54c\uace0\ub9ac\uc998\uc73c\ub85c, \uac1d\uccb4\ub97c \ud0d0\uc9c0\ud560 \ub54c \uc774\ubbf8\uc9c0\ub97c \uc5ec\ub7ec \uac1c\uc758 \uadf8\ub9ac\ub4dc\ub85c \ub098\ub204\uace0
\n \uac01 \uadf8\ub9ac\ub4dc\uc5d0\uc11c \uac1d\uccb4\uc758 \uacbd\uacc4 \uc0c1\uc790(bounding box)\uc640 \ud074\ub798\uc2a4 \ud655\ub960\uc744 \uc608\uce21\ud569\ub2c8\ub2e4.
\n \uc774\ub294 \uc18d\ub3c4\uac00 \ube60\ub974\uba70 \ub192\uc740 \uc815\ud655\ub3c4\ub97c \uc790\ub791\ud558\uae30 \ub54c\ubb38\uc5d0 \uc2e4\uc2dc\uac04 \uac1d\uccb4 \uac80\ucd9c\uc5d0 \uc801\ud569\ud569\ub2c8\ub2e4.\n <\/p>\n

1.1 YOLO\uc758 \uc791\ub3d9 \uc6d0\ub9ac<\/h3>\n

YOLO\ub294 \uc774\ubbf8\uc9c0 \uc804\uccb4\ub97c \ud55c \ubc88\uc5d0 \ucc98\ub9ac\ud558\ub294 \ubc29\uc2dd\uc73c\ub85c \uc791\ub3d9\ud569\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n

    \n
  1. \uc774\ubbf8\uc9c0\ub97c S x S \uadf8\ub9ac\ub4dc\ub85c \ub098\ub208\ub2e4.<\/li>\n
  2. \uac01 \uadf8\ub9ac\ub4dc \uc140\uc740 \uadf8\ub9ac\ub4dc \ub0b4\uc5d0\uc11c \uac1d\uccb4\uac00 \uc788\ub294 \uacbd\uc6b0, \ud574\ub2f9 \uac1d\uccb4\uc758 \uacbd\uacc4 \uc0c1\uc790\uc640 \ud074\ub798\uc2a4 \ud655\ub960\uc744 \uc608\uce21\ud55c\ub2e4.<\/li>\n
  3. \uc608\uce21\ub41c \uacbd\uacc4 \uc0c1\uc790 \ubc0f \ud074\ub798\uc2a4 \ud655\ub960\uc744 \ubc14\ud0d5\uc73c\ub85c \ucd5c\uc885 \uac1d\uccb4 \uac80\ucd9c \uacb0\uacfc\ub97c \uc0dd\uc131\ud55c\ub2e4.<\/li>\n<\/ol>\n

    \uae30\ubcf8\uc801\uc73c\ub85c YOLO\ub294 CNN(Convolutional Neural Network)\uc744 \uae30\ubc18\uc73c\ub85c \ud558\uc5ec \uac1d\uccb4 \ud0d0\uc9c0\ub97c \uc218\ud589\ud558\uba70,
    \n \uac01\uac01\uc758 \uadf8\ub9ac\ub4dc \uc140\uc5d0\uc11c \uac1d\uccb4\uac00 \uc874\uc7ac\ud560 \ud655\ub960\uacfc \ud568\uaed8 \uacbd\uacc4 \uc0c1\uc790\ub97c \uc608\uce21\ud569\ub2c8\ub2e4.<\/p>\n

    2. YOLO \ubaa8\ub378 \ub2e4\uc6b4\ub85c\ub4dc<\/h2>\n

    \n YOLO \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud558\uae30 \uc704\ud574\uc11c\ub294 \ud559\uc2b5\ub41c \uac00\uc911\uce58 \ud30c\uc77c\uc774 \ud544\uc694\ud569\ub2c8\ub2e4.
    \n \uc5ec\uae30\uc11c\ub294 YOLOv3 \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud560 \uac83\uc785\ub2c8\ub2e4. YOLOv3\uc758 \uac00\uc911\uce58\ub97c \ub2e4\uc6b4\ub85c\ub4dc\ud558\ub824\uba74 \uc544\ub798\uc758 \uba85\ub839\uc5b4\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n

    wget https:\/\/pjreddie.com\/media\/files\/yolov3.weights<\/code><\/pre>\n
    3. OpenCV \uc124\uce58<\/h2>\n
    \n        YOLO\ub97c OpenCV\uc640 \ud568\uaed8 \uc0ac\uc6a9\ud558\uae30 \uc704\ud574 OpenCV \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc124\uce58\ud574\uc57c \ud569\ub2c8\ub2e4.
    \n        Python\uc5d0\uc11c OpenCV\ub97c \uc124\uce58\ud558\ub824\uba74 pip \uba85\ub839\uc5b4\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n    <\/p>\n
    pip install opencv-python<\/code><\/pre>\n
    4. YOLO\uc640 OpenCV \ud1b5\ud569<\/h2>\n
    \n        \uc774\uc81c YOLO \ubaa8\ub378\uacfc OpenCV\ub97c \ud1b5\ud569\ud558\uc5ec \uc2e4\uc2dc\uac04 \uac1d\uccb4 \uac80\ucd9c\uc744 \uc218\ud589\ud558\ub294 \ucf54\ub4dc\ub97c \uc791\uc131\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.
    \n        \uc544\ub798 \ucf54\ub4dc\ub294 \uc6f9\ucea0\uc5d0\uc11c \ud504\ub808\uc784\uc744 \uc77d\uace0 YOLO\ub97c \uc0ac\uc6a9\ud558\uc5ec \uac1d\uccb4\ub97c \uac80\ucd9c\ud558\ub294 \uc608\uc81c\uc785\ub2c8\ub2e4.\n    <\/p>\n
    import cv2\nimport numpy as np\n\n# YOLO \ubaa8\ub378 \uc124\uc815 \ud30c\uc77c \uacbd\ub85c\nconfigPath = \"yolov3.cfg\"\nweightsPath = \"yolov3.weights\"\n\n# \ud074\ub798\uc2a4 \uc774\ub984\uc744 \uac00\uc838\uc635\ub2c8\ub2e4.\nwith open(\"coco.names\", \"r\") as f:\n    classes = [line.strip() for line in f.readlines()]\n\n# YOLO \ub124\ud2b8\uc6cc\ud06c\ub97c \ub85c\ub4dc\ud569\ub2c8\ub2e4.\nnet = cv2.dnn.readNet(weightsPath, configPath)\n\n# \uc6f9\ucea0\uc744 \uc624\ud508\ud569\ub2c8\ub2e4.\ncap = cv2.VideoCapture(0)\n\nwhile True:\n    # \uc6f9\ucea0\uc5d0\uc11c \ud504\ub808\uc784\uc744 \uc77d\uc5b4 \uc635\ub2c8\ub2e4.\n    _, frame = cap.read()\n    height, width, _ = frame.shape\n\n    # YOLO\uc758 \uc785\ub825\uc73c\ub85c \uc0ac\uc6a9\ud560 blob\uc744 \uc0dd\uc131\ud569\ub2c8\ub2e4.\n    blob = cv2.dnn.blobFromImage(frame, 0.00392, (416, 416), (0, 0, 0), True, crop=False)\n    net.setInput(blob)\n\n    # \ub124\ud2b8\uc6cc\ud06c\uc758 \ucd9c\ub825\uc744 \uac00\uc838\uc635\ub2c8\ub2e4.\n    layerNames = net.getLayerNames()\n    outputLayers = [layerNames[i[0] - 1] for i in net.getUnconnectedOutLayers()]\n    outputs = net.forward(outputLayers)\n\n    boxes = []\n    confidences = []\n    classIDs = []\n\n    # \ucd9c\ub825\uc744 \ubd84\uc11d\ud558\uc5ec \uac1d\uccb4 \uc815\ubcf4\ub97c \ud68d\ub4dd\ud569\ub2c8\ub2e4.\n    for output in outputs:\n        for detection in output:\n            scores = detection[5:]\n            classID = np.argmax(scores)\n            confidence = scores[classID]\n\n            if confidence > 0.5:  # \uc2e0\ub8b0\ub3c4\uac00 0.5 \uc774\uc0c1\uc778 \uacbd\uc6b0\ub9cc \ucc98\ub9ac\n                # \uac1d\uccb4\uc758 \uc704\uce58\ub97c \uacc4\uc0b0\ud569\ub2c8\ub2e4.\n                centerX = int(detection[0] * width)\n                centerY = int(detection[1] * height)\n                w = int(detection[2] * width)\n                h = int(detection[3] * height)\n\n                # \uc9c1\uc0ac\uac01\ud615\uc758 \uc88c\uc0c1\ub2e8 \uc88c\ud45c\ub97c \uacc4\uc0b0\ud569\ub2c8\ub2e4.\n                x = int(centerX - w \/ 2)\n                y = int(centerY - h \/ 2)\n\n                boxes.append([x, y, w, h])\n                confidences.append(float(confidence))\n                classIDs.append(classID)\n\n    # \ube44\ucd5c\ub300 \uc5b5\uc81c(Non-Maximum Suppression)\ub97c \uc801\uc6a9\ud569\ub2c8\ub2e4.\n    indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)\n\n    # \ucd5c\uc885\uc801\uc73c\ub85c \uacb0\uacfc\ub97c \ud654\uba74\uc5d0 \ud45c\uc2dc\ud569\ub2c8\ub2e4.\n    for i in range(len(boxes)):\n        if i in indexes:\n            x, y, w, h = boxes[i]\n            label = str(classes[classIDs[i]])\n            cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)\n            cv2.putText(frame, label, (x, y + 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)\n\n    # \uacb0\uacfc \ud504\ub808\uc784\uc744 \ud654\uba74\uc5d0 \ud45c\uc2dc\ud569\ub2c8\ub2e4.\n    cv2.imshow(\"YOLO Object Detection\", frame)\n\n    # 'q' \ud0a4\ub97c \ub20c\ub7ec \uc885\ub8cc\ud569\ub2c8\ub2e4.\n    if cv2.waitKey(1) & 0xFF == ord('q'):\n        break\n\n# \uc6f9\ucea0\uacfc \ucc3d\uc744 \uc885\ub8cc\ud569\ub2c8\ub2e4.\ncap.release()\ncv2.destroyAllWindows()\n<\/code><\/pre>\n
    5. \ucf54\ub4dc \uc124\uba85<\/h2>\n
    \n        \uc704 \ucf54\ub4dc\ub294 \uc6f9\ucea0\uc5d0\uc11c \uc2e4\uc2dc\uac04\uc73c\ub85c \uac1d\uccb4\ub97c \ud0d0\uc9c0\ud558\ub294 \uc804\uccb4 \uc560\ud50c\ub9ac\ucf00\uc774\uc158\uc785\ub2c8\ub2e4.
    \n        \ucf54\ub4dc\uc758 \uc8fc\uc694 \ubd80\ubd84\uc744 \uc0b4\ud3b4\ubcf4\uba74 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:\n    <\/p>\n