yolov3学习之image_preporcess

原代码地址：GitHub - YunYang1994/tensorflow-yolov3

1. 先贴一下原image_preporcess函数

def image_preporcess(image, target_size, gt_boxes=None):

    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB).astype(np.float32)

    ih, iw    = target_size
    h,  w, _  = image.shape

    scale = min(iw/w, ih/h)
    nw, nh  = int(scale * w), int(scale * h)
    image_resized = cv2.resize(image, (nw, nh))

    image_paded = np.full(shape=[ih, iw, 3], fill_value=128.0)
    dw, dh = (iw - nw) // 2, (ih-nh) // 2
    image_paded[dh:nh+dh, dw:nw+dw, :] = image_resized
    image_paded = image_paded / 255.

    if gt_boxes is None:
        return image_paded

    else:
        gt_boxes[:, [0, 2]] = gt_boxes[:, [0, 2]] * scale + dw
        gt_boxes[:, [1, 3]] = gt_boxes[:, [1, 3]] * scale + dh
        return image_paded, gt_boxes

2. 测试一下调用结果

在网上找了张风景图，将它的大小调成400(height)X800(width)：

采用如下代码处理上述风景图片，把输出目标大小设置为[416,416]：

if __name__ == "__main__":
    img = cv2.imread('./landscape.jpg')
    print('===> img.shape: ', img.shape) # height=400, width=800
    target_size = [416, 416]
    processed_img = image_preporcess(img, target_size)
    print('===> processed_img.shape: ', processed_img.shape)
    
    import matplotlib.pyplot as plt
    plt.imshow(processed_img)
    plt.show()

打印原图和处理后图片的大小：

 处理后的图片：

 3.代码理解

def image_preporcess(image, target_size, gt_boxes=None):
    # 采用opencv读取的图片是BGR格式的，需要对其进行转化为RGB格式
    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB).astype(np.float32)

    ih, iw    = target_size
    h,  w, _  = image.shape
    
    # 设目标大小为416X416，通俗来讲，这一步骤就是要将原图缩放到正好能放
    # 进416X416的框里，最长边调整为416，短的一边根据原图的高宽比进行调整
    scale = min(iw/w, ih/h)
    nw, nh  = int(scale * w), int(scale * h)
    image_resized = cv2.resize(image, (nw, nh))

    # 先创建一个416X416的框，所有像素用设定好的像素进行填充。
    image_paded = np.full(shape=[ih, iw, 3], fill_value=128.0)

    # 为了将缩放后的图片放到416X416这个框的中心位置，计算需要偏移的量dw, dh
    dw, dh = (iw - nw) // 2, (ih-nh) // 2

    # 在image_paded这个416X416的框里找到相应的位置，把缩放好的图片放进去
    image_paded[dh:nh+dh, dw:nw+dw, :] = image_resized

    # 最后对图片进行归一化处理
    image_paded = image_paded / 255.