資料增強，對圖檔的crop，以及對關鍵點标注的對應縮放

以以下圖檔(000000391895.jpg)為例子，可視化圖檔裁剪過程

對圖檔裁剪的關鍵在于如何讓關鍵點标注在裁剪前後同步進行坐标點的轉換。要裁剪的區域為gt_box，但需要對gt_box進行一定的擴充，首先算出中心點objcenter，再根據中心點進行高和寬的縮放，然後對圖檔進行一定的裁剪。

其實我還是有點沒搞懂代碼，為什麼有個crop_size和min_shape出來。

詳細可參考此部落格

if __name__ == '__main__':
    import matplotlib.patches as patches


    def augmentationCropImage(img, bbox, joints=None):
        is_train = True
        height, width = 256, 192
        bbox = np.array(bbox).reshape(4, ).astype(np.float32)
        add = max(img.shape[0], img.shape[1])
        mean_value = np.array([122.7717, 115.9465, 102.9801])
        bimg = cv2.copyMakeBorder(img, add, add, add, add, borderType=cv2.BORDER_CONSTANT, value=mean_value.tolist())
        objcenter = np.array([(bbox[0] + bbox[2]) / 2., (bbox[1] + bbox[3]) / 2.])
        bbox += add
        objcenter += add
        if is_train:
            joints[:, :2] += add
            inds = np.where(joints[:, -1] == 0)
            joints[inds, :2] = -1000000  # avoid influencing by data processing
        crop_width = (bbox[2] - bbox[0]) * (1 + 0.1 * 2)
        crop_height = (bbox[3] - bbox[1]) * (1 + 0.15 * 2)
        if is_train:
            crop_width = crop_width * (1 + 0.25)
            crop_height = crop_height * (1 + 0.25)
        if crop_height / height > crop_width / width:
            crop_size = crop_height
            min_shape = height
        else:
            crop_size = crop_width
            min_shape = width

        crop_size = min(crop_size, objcenter[0] / width * min_shape * 2. - 1.)
        crop_size = min(crop_size, (bimg.shape[1] - objcenter[0]) / width * min_shape * 2. - 1)
        crop_size = min(crop_size, objcenter[1] / height * min_shape * 2. - 1.)
        crop_size = min(crop_size, (bimg.shape[0] - objcenter[1]) / height * min_shape * 2. - 1)

        min_x = int(objcenter[0] - crop_size / 2. / min_shape * width)
        max_x = int(objcenter[0] + crop_size / 2. / min_shape * width)
        min_y = int(objcenter[1] - crop_size / 2. / min_shape * height)
        max_y = int(objcenter[1] + crop_size / 2. / min_shape * height)

        x_ratio = float(width) / (max_x - min_x)
        y_ratio = float(height) / (max_y - min_y)

        if is_train:
            joints[:, 0] = joints[:, 0] - min_x
            joints[:, 1] = joints[:, 1] - min_y

            joints[:, 0] *= x_ratio
            joints[:, 1] *= y_ratio
            label = joints[:, :2].copy()
            valid = joints[:, 2].copy()

        img = cv2.resize(bimg[min_y:max_y, min_x:max_x, :], (width, height))
        details = np.asarray([min_x - add, min_y - add, max_x - add, max_y - add]).astype(np.float)

        if is_train:
            return img, joints, details
        else:
            return img, details


    num_class = 17
    a =  {'unit': {'num_keypoints': 14, 'keypoints': [368, 61, 1, 369, 52, 2, 0, 0, 0, 382, 48, 2, 0, 0, 0, 368, 84, 2, 435, 81, 2, 362, 125, 2, 446, 125, 2, 360, 153, 2, 0, 0, 0, 397, 167, 1, 439, 166, 1, 369, 193, 2, 461, 234, 2, 361, 246, 2, 474, 287, 2], 'GT_bbox': [339, 22, 493, 322]}, 'imgInfo': {'imgID': 391895, 'img_paths': '000000391895.jpg'}, 'operation': 0}
    image_name = a['imgInfo']['img_paths']
    points = np.array(a['unit']['keypoints']).reshape(num_class, 3).astype(np.float32)
    gt_bbox = a['unit']['GT_bbox']
    points = points.flatten()
    x = points[0::3]
    y = points[1::3]
    v = points[2::3]
    plt.figure()
    c = (np.random.random((1, 3)) * 0.6 + 0.4).tolist()[0]
    plt.plot(x[v > 0], y[v > 0], 'o', markersize=10, markerfacecolor=c, markeredgecolor='k', markeredgewidth=2)

    x = gt_bbox[0]
    y = gt_bbox[1]
    width = gt_bbox[2] - gt_bbox[0]
    height = gt_bbox[3] - gt_bbox[1]
    currentAxis = plt.gca()
    rect = patches.Rectangle((x, y), width, height, linewidth=10, edgecolor='r', facecolor='none')
    currentAxis.add_patch(rect)
    img =  cv2.imread(image_name)

    plt.imshow(img)
    plt.show()


    points = np.array(a['unit']['keypoints']).reshape(num_class, 3).astype(np.float32)
    image, points, details = augmentationCropImage(img, gt_bbox, points)


    # 畫矩形
    x = details[0]
    y = details[1]
    width = details[2] - details[0]
    height = details[3] - details[1]
    currentAxis = plt.gca()
    rect = patches.Rectangle((x, y), width, height, linewidth=10, edgecolor='r', facecolor='none')
    currentAxis.add_patch(rect)
    # 畫關鍵點
    points = points.flatten()
    x = points[0::3]
    y = points[1::3]
    v = points[2::3]
    plt.plot(x, y, 'o', markersize=10, markerfacecolor='red', markeredgecolor='k', markeredgewidth=2)
    # 展示圖檔
    plt.imshow(image)
    plt.show()