前言
數據是深度學習的核心，大部分論文裡都會提到data-driven這個詞，也就是
數據集的均值和標準差
<code>def compute_mean_and_std(dataset): # 輸入為PyTorch的dataset，即數據集，輸出為對應數據集均值和標準差 # 均值 mean_r = 0 mean_g = 0 mean_b = 0 for img, _ in dataset: img = np.asarray(img) # 將 PIL Image 改變成numpy的數組類型 mean_b += np.mean(img[:, :, 0]) mean_g += np.mean(img[:, :, 1]) mean_r += np.mean(img[:, :, 2]) mean_b /= len(dataset) mean_g /= len(dataset) mean_r /= len(dataset) diff_r = 0 diff_g = 0 diff_b = 0 N = 0 for img, _ in dataset: img = np.asarray(img) diff_b += np.sum(np.power(img[:, :, 0] - mean_b, 2)) diff_g += np.sum(np.power(img[:, :, 1] - mean_g, 2)) diff_r += np.sum(np.power(img[:, :, 2] - mean_r, 2)) N += np.prod(img[:, :, 0].shape) std_b = np.sqrt(diff_b / N) std_g = np.sqrt(diff_g / N) std_r = np.sqrt(diff_r / N) mean = (mean_b.item() / 255.0, mean_g.item() / 255.0, mean_r.item() / 255.0) std = (std_b.item() / 255.0, std_g.item() / 255.0, std_r.item() / 255.0) return mean, std/<code>
常用訓練和驗證數據預處理
ToTensor 會將 PIL.Image 或形狀為 H×W×D，數值範圍為 [0, 255] 的numpy數組轉換為形狀為 D×H×W，數值範圍為 [0.0, 1.0] 的 torch.Tensor。
<code>train_transform = torchvision.transforms.Compose([ torchvision.transforms.RandomResizedCrop(size=224, scale=(0.08, 1.0)), torchvision.transforms.RandomHorizontalFlip(), torchvision.transforms.ToTensor(), torchvision.transforms.Normalize(mean=(0.485, 0.456, 0.406)std=(0.229, 0.224, 0.225)), ]) val_transform = torchvision.transforms.Compose([ torchvision.transforms.Resize(256), torchvision.transforms.CenterCrop(224), torchvision.transforms.ToTensor(), torchvision.transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)), ])/<code>
視頻數據
<code>import cv2 video = cv2.VideoCapture(mp4_path) height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT)) width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH)) num_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT)) fps = int(video.get(cv2.CAP_PROP_FPS)) video.release()/<code>
未完待續，持續更新！