\uc800\uc790: \uc870\uad11\ud615<\/p>\n
\uc791\uc131\uc77c: 2024\ub144 11\uc6d4 26\uc77c<\/p>\n<\/header>\n \n OpenCV(Open Source Computer Vision Library)\ub294 \ucef4\ud4e8\ud130 \ube44\uc804 \ubc0f \uba38\uc2e0\ub7ec\ub2dd\uc744 \uc704\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\ub85c, \uc774 \uac15\uc88c\ub97c \uc9c4\ud589\ud558\uae30 \uc704\ud574 \ud544\uc694\ud55c \uc694\uad6c \uc0ac\ud56d\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n Python\uacfc \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc124\uce58\ud558\ub824\uba74 \ub2e4\uc74c\uacfc \uac19\uc740 \uba85\ub839\uc5b4\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4:<\/p>\n \n \uc6b0\uc120 PyTorch\ub97c \uc0ac\uc6a9\ud558\uc5ec \uac04\ub2e8\ud55c CNN \ubaa8\ub378\uc744 \ud559\uc2b5\uc2dc\ud0b5\ub2c8\ub2e4. \uc5ec\uae30\uc11c\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790 \uc774\ubbf8\uc9c0\ub97c \ubd84\ub958\ud558\ub294 \ubaa8\ub378\uc744 \ub9cc\ub4e4\uc5b4 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n \n \uc704 \ucf54\ub4dc\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \ub85c\ub4dc\ud558\uc5ec \uac04\ub2e8\ud55c CNN \ubaa8\ub378\uc744 \uc815\uc758\ud558\uace0 \ud559\uc2b5\uc2dc\ud0a4\ub294 \ucf54\ub4dc\uc785\ub2c8\ub2e4. \n \uc800\uc7a5\ud55c PyTorch \ubaa8\ub378\uc744 OpenCV\uc5d0\uc11c \uc811\uadfc\ud558\ub824\uba74, \uba3c\uc800 \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\uace0 OpenCV \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud574\uc57c \ud569\ub2c8\ub2e4. \n \uc704 \ucf54\ub4dc\ub294 \ubaa8\ub378\uc744 ONNX \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud558\ub294 \uacfc\uc815\uc785\ub2c8\ub2e4. \uc774\uc81c OpenCV\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc774 ONNX \ubaa8\ub378\uc744 \ub85c\ub4dc\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n<\/section>\n \n OpenCV\uc5d0\uc11c ONNX \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\uace0 \uc2e4\ud589\ud558\ub294 \ubc29\ubc95\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4. OpenCV\uc758 dnn \ubaa8\ub4c8\uc744 \uc0ac\uc6a9\ud558\uc5ec \n \uc704 \ucf54\ub4dc\ub294 OpenCV\ub97c \uc0ac\uc6a9\ud558\uc5ec ONNX \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\uace0, \uc8fc\uc5b4\uc9c4 \uc774\ubbf8\uc9c0\ub97c \uc804\ucc98\ub9ac\ud55c \ud6c4 \ubaa8\ub378\uc744 \ud1b5\ud574 \n OpenCV\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc6f9\ucea0\uc5d0\uc11c \uc2e4\uc2dc\uac04\uc73c\ub85c \uc774\ubbf8\uc9c0\ub97c \ucc98\ub9ac\ud558\uace0 \uc608\uce21\ud558\ub294 \ubc29\ubc95\ub3c4 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \n \uc704 \ucf54\ub4dc\ub294 \uc6f9\ucea0\uc73c\ub85c\ubd80\ud130 \ud504\ub808\uc784\uc744 \uac00\uc838\uc624\uace0 \uac01 \ud504\ub808\uc784\uc5d0 \ub300\ud574 \uc804\ucc98\ub9ac\ub97c \uc218\ud589\ud558\uc5ec \ubaa8\ub378\uc744 \ud1b5\ud574 \n \uc774\ubc88 \uac15\uc88c\uc5d0\uc11c\ub294 PyTorch\ub85c \ud559\uc2b5\ud55c CNN \ubaa8\ub378\uc744 OpenCV\uc5d0\uc11c \ub85c\ub4dc\ud558\uc5ec \uc2e4\ud589\ud558\ub294 \ubc29\ubc95\uc744 \uc54c\uc544\ubcf4\uc558\uc2b5\ub2c8\ub2e4.1. \uc11c\ub860<\/h2>\n
\n \ub2e4\uc591\ud55c \uc774\ubbf8\uc9c0 \ubc0f \ube44\ub514\uc624 \ucc98\ub9ac \uae30\uc220\uc744 \uc81c\uacf5\ud569\ub2c8\ub2e4. PyTorch\ub294 \uba38\uc2e0\ub7ec\ub2dd \ub77c\uc774\ube0c\ub7ec\ub9ac\ub85c, \ud2b9\ud788 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc758
\n \uac1c\ubc1c\uacfc \ud559\uc2b5\uc5d0 \ub9ce\uc774 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. \ubcf8 \uac15\uc88c\uc5d0\uc11c\ub294 PyTorch\ub85c \ud559\uc2b5\ud55c \ubaa8\ub378\uc744 OpenCV\ub97c \ud1b5\ud574 \ub85c\ub4dc\ud558\uc5ec
\n \uc2e4\uc2dc\uac04\uc73c\ub85c \uc2e4\ud589\ud558\ub294 \ubc29\ubc95\uc5d0 \ub300\ud574 \uc124\uba85\ud558\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n<\/section>\n2. \uc694\uad6c \uc0ac\ud56d<\/h2>\n
\n
pip install opencv-python torchvision torch numpy<\/code><\/pre>\n<\/section>\n3. PyTorch \ubaa8\ub378 \ud559\uc2b5\ud558\uae30<\/h2>\n
\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\n\n# \ub370\uc774\ud130\uc14b \ub85c\ub4dc \ubc0f \uc804\ucc98\ub9ac\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\ntrain_dataset = datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ntrain_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)\n\n# \ubaa8\ub378 \uc815\uc758\nclass SimpleCNN(nn.Module):\n def __init__(self):\n super(SimpleCNN, self).__init__()\n self.conv1 = nn.Conv2d(1, 32, kernel_size=3)\n self.pool = nn.MaxPool2d(kernel_size=2)\n self.conv2 = nn.Conv2d(32, 64, kernel_size=3)\n self.fc1 = nn.Linear(64 * 6 * 6, 128)\n self.fc2 = nn.Linear(128, 10)\n \n def forward(self, x):\n x = self.pool(torch.relu(self.conv1(x)))\n x = self.pool(torch.relu(self.conv2(x)))\n x = x.view(-1, 64 * 6 * 6)\n x = torch.relu(self.fc1(x))\n return self.fc2(x)\n\n# \ubaa8\ub378 \uc778\uc2a4\ud134\uc2a4 \uc0dd\uc131 \ubc0f \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998 \uc815\uc758\nmodel = SimpleCNN()\ncriterion = nn.CrossEntropyLoss()\noptimizer = optim.Adam(model.parameters(), lr=0.001)\n\n# \ubaa8\ub378 \ud559\uc2b5\nmodel.train()\nfor epoch in range(5): # 5 \uc5d0\ud3ec\ud06c \ub3d9\uc548 \ud559\uc2b5\n for images, labels in train_loader:\n optimizer.zero_grad()\n outputs = model(images)\n loss = criterion(outputs, labels)\n loss.backward()\n optimizer.step()\n \n print(f'Epoch [{epoch+1}\/5], Loss: {loss.item():.4f}')\n\n# \ubaa8\ub378 \uc800\uc7a5\ntorch.save(model.state_dict(), 'mnist_cnn.pth')\n <\/code><\/pre>\n
\n \ubaa8\ub378\uc774 \ud559\uc2b5\ud55c \ud6c4, ‘mnist_cnn.pth’ \ud30c\uc77c\ub85c \ubaa8\ub378\uc744 \uc800\uc7a5\ud569\ub2c8\ub2e4.\n <\/p>\n<\/section>\n4. OpenCV\uc5d0\uc11c PyTorch \ubaa8\ub378 \ub85c\ub4dc\ud558\uae30<\/h2>\n
\n OpenCV\uc5d0\uc11c \uc0ac\uc6a9\ud558\ub824\uba74 ONNX(Open Neural Network Exchange) \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud574\uc57c \ud569\ub2c8\ub2e4. \uc544\ub798 \ucf54\ub4dc\ub97c
\n \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc744 ONNX \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud569\ub2c8\ub2e4.\n <\/p>\n\ndummy_input = torch.randn(1, 1, 28, 28) # MNIST\uc758 \uc774\ubbf8\uc9c0 \ud06c\uae30\ntorch.onnx.export(model, dummy_input, 'mnist_cnn.onnx')\n <\/code><\/pre>\n5. OpenCV\ub97c \uc0ac\uc6a9\ud558\uc5ec ONNX \ubaa8\ub378 \uc2e4\ud589\ud558\uae30<\/h2>\n
\n \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\uace0 \uc774\ubbf8\uc9c0\ub97c \uc785\ub825\uc73c\ub85c \uc0ac\uc6a9\ud558\uc5ec \uc608\uce21\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n\nimport cv2\nimport numpy as np\n\n# \ubaa8\ub378 \ub85c\ub4dc\nnet = cv2.dnn.readNetFromONNX('mnist_cnn.onnx')\n\n# \uc774\ubbf8\uc9c0 \uc804\ucc98\ub9ac\nimage = cv2.imread('test_image.png', cv2.IMREAD_GRAYSCALE)\nimage = cv2.resize(image, (28, 28))\nimage = image.astype(np.float32) \/ 255.0\nimage = np.expand_dims(image, axis=0)\nimage = np.expand_dims(image, axis=0)\nimage = np.array(image, dtype=np.float32)\n\n# \ubaa8\ub378\uc5d0 \uc785\ub825\nblob = cv2.dnn.blobFromImage(image)\nnet.setInput(blob)\noutput = net.forward()\n\n# \uacb0\uacfc \ucd9c\ub825\npredicted_class = np.argmax(output, axis=1)\nprint(f'Predicted class: {predicted_class[0]}')\n <\/code><\/pre>\n
\n \uc608\uce21 \uacb0\uacfc\ub97c \ucd9c\ub825\ud569\ub2c8\ub2e4. \uc774\ubbf8\uc9c0\uac00 MNIST \ub370\uc774\ud130\uc14b\uc758 \ud06c\uae30\uc5d0 \ub9de\uac8c \uc804\ucc98\ub9ac\ub418\uc5c8\ub294\uc9c0 \ud655\uc778\ud558\uc138\uc694.\n <\/p>\n<\/section>\n6. \uc2e4\uc2dc\uac04 \uc774\ubbf8\uc9c0 \ucc98\ub9ac<\/h2>\n
\n \uc544\ub798 \ucf54\ub4dc\ub294 \uc6f9\ucea0\uc5d0\uc11c \uc774\ubbf8\uc9c0\ub97c \ucea1\ucc98\ud558\uace0, \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790\ub97c \uc778\uc2dd\ud558\ub294 \uc608\uc81c\uc785\ub2c8\ub2e4.\n <\/p>\n\ncap = cv2.VideoCapture(0) # \uc6f9\ucea0 \ucea1\ucc98 \uc2dc\uc791\n\nwhile True:\n ret, frame = cap.read()\n if not ret:\n break\n \n # \uc774\ubbf8\uc9c0 \uc804\ucc98\ub9ac\n gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)\n resized = cv2.resize(gray, (28, 28))\n normed = resized.astype(np.float32) \/ 255.0\n blob = cv2.dnn.blobFromImage(normed)\n \n # \ubaa8\ub378\uc5d0 \uc785\ub825\n net.setInput(blob)\n output = net.forward()\n predicted_class = np.argmax(output, axis=1)[0]\n \n # \uc608\uce21 \uacb0\uacfc \ud45c\uc2dc\n cv2.putText(frame, f'Predicted: {predicted_class}', (10, 30), cv2.FONT_HERSHEY_SIMPLEX, \n 1, (0, 255, 0), 2, cv2.LINE_AA)\n \n cv2.imshow('Webcam', frame)\n \n if cv2.waitKey(1) & 0xFF == ord('q'):\n break\n\ncap.release()\ncv2.destroyAllWindows()\n <\/code><\/pre>\n
\n \uc608\uce21\uc744 \uc218\ud589\ud558\uace0, \uc608\uce21\ub41c \uc22b\uc790\ub97c \ud654\uba74\uc5d0 \ud45c\uc2dc\ud569\ub2c8\ub2e4.
\n ‘q’ \ud0a4\ub97c \ub20c\ub7ec\uc11c \uc885\ub8cc\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n<\/section>\n7. \uacb0\ub860<\/h2>\n
\n ONNX \ud615\uc2dd\uc73c\ub85c \ubaa8\ub378\uc744 \ubcc0\ud658\ud558\uace0, OpenCV\uc758 DNN \ubaa8\ub4c8\uc744 \ud65c\uc6a9\ud558\uc5ec \uc2e4\uc2dc\uac04 \uc774\ubbf8\uc9c0 \ucc98\ub9ac\uc5d0 \ud65c\uc6a9\ud560 \uc218 \uc788\uc74c\uc744
\n \ubc30\uc6e0\uc2b5\ub2c8\ub2e4. \uc774\ub7ec\ud55c \ubc29\uc2dd\uc73c\ub85c \uae30\uc874\uc758 \ub2e4\uc591\ud55c \uba38\uc2e0\ub7ec\ub2dd \ubaa8\ub378\uc744 OpenCV\uc640 \ud1b5\ud569\ud558\uc5ec \uc774\ubbf8\uc9c0\ub098 \ube44\ub514\uc624 \ucc98\ub9ac
\n \uc791\uc5c5\uc5d0 \ud65c\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ud5a5\ud6c4 \ub354 \ubcf5\uc7a1\ud55c \ubaa8\ub378 \ubc0f \ub2e4\uc591\ud55c \ub370\uc774\ud130\uc14b\uc73c\ub85c\uc758 \ud655\uc7a5\uc744 \uace0\ub824\ud574 \ubcf4\uc138\uc694.\n <\/p>\n<\/section>\n