您好，首先感谢您的工作，这给了我很大的启发。现在我也在考虑这个方向的相关工作。在处理DroneVehicle数据集时看到了一处我不理解的地方，在/workspace/DPAL/dataset/preprocess_DroneVehicle.py文件中，您处理DroneVehicle这个数据集时仅仅只删去了上下左右4个像素点的白边，但是DroneVehicle数据集上下左右共有100个像素点的白边，为什么不全部删去呢？源代码如下：
def process_images_and_annotations(image_folder, new_image_folder, json_path, new_json_path):
# 处理图像，拆去上下左右4%的白边
for filename in os.listdir(image_folder):
if filename.endswith(('.jpg', '.png', '.jpeg')):
image_path = os.path.join(image_folder, filename)
img = Image.open(image_path)
new_img = img.crop((4, 4, img.width - 4, img.height - 4))
new_image_path = os.path.join(new_image_folder, filename)
new_img.save(new_image_path)
# 处理标注
with open(json_path, 'r') as f:
data = json.load(f)
for annotation in data['annotations']:
bbox = annotation['bbox']
x, y, w, h = bbox
new_x = x - 4 if x > 4 else 0
new_y = y - 4 if y > 4 else 0
new_w = w if new_x + w < 832 else 832 - new_x
new_h = h if new_y + h < 704 else 704 - new_y
annotation['bbox'] = [new_x, new_y, new_w, new_h]
with open(new_json_path, 'w') as f:
json.dump(data, f)
您好，首先感谢您的工作，这给了我很大的启发。现在我也在考虑这个方向的相关工作。在处理DroneVehicle数据集时看到了一处我不理解的地方，在/workspace/DPAL/dataset/preprocess_DroneVehicle.py文件中，您处理DroneVehicle这个数据集时仅仅只删去了上下左右4个像素点的白边，但是DroneVehicle数据集上下左右共有100个像素点的白边，为什么不全部删去呢？源代码如下：
def process_images_and_annotations(image_folder, new_image_folder, json_path, new_json_path):
# 处理图像，拆去上下左右4%的白边
for filename in os.listdir(image_folder):
if filename.endswith(('.jpg', '.png', '.jpeg')):
image_path = os.path.join(image_folder, filename)
img = Image.open(image_path)
new_img = img.crop((4, 4, img.width - 4, img.height - 4))
new_image_path = os.path.join(new_image_folder, filename)
new_img.save(new_image_path)