paddleocr Out of memory error

[kennyLSN]

问题描述 / Problem Description

paddleocr 做长图的ocr识别，det_limit_side_len=20000，10G 显存，Out of memory error on

运行环境 / Runtime Environment

• OS:
• Paddle: paddlepaddle-gpu==2.6.0
• PaddleOCR:

NVIDIA-SMI 470.57.02 Driver Version: 470.57.02 CUDA Version: 11.8 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA A10 On | 00000000:00:09.0 Off | 0 |
| 0% 52C P0 64W / 150W | 10262MiB / 10332MiB | 1% Default |
| | | N/A |

复现代码 / Reproduction Code

完整报错 / Complete Error Message

7 phi::DnnWorkspaceHandle::RunFunc(std::function<void (void*)> const&, unsigned long)
8 phi::DnnWorkspaceHandle::ReallocWorkspace(unsigned long)
9 paddle::memory::allocation::Allocator::Allocate(unsigned long)
10 paddle::memory::allocation::StatAllocator::AllocateImpl(unsigned long)
11 paddle::memory::allocation::Allocator::Allocate(unsigned long)
12 paddle::memory::allocation::Allocator::Allocate(unsigned long)
13 paddle::memory::allocation::Allocator::Allocate(unsigned long)
14 paddle::memory::allocation::Allocator::Allocate(unsigned long)
15 paddle::memory::allocation::CUDAAllocator::AllocateImpl(unsigned long)
16 std::string phi::enforce::GetCompleteTraceBackString<std::string >(std::string&&, char const*, int)
17 phi::enforce::GetCurrentTraceBackStringabi:cxx11

---

Error Message Summary:

ResourceExhaustedError:

Out of memory error on GPU 0. Cannot allocate 128.000000MB memory on GPU 0, 10.021484GB memory has been allocated and available memory is only 70.000000MB.

Please check whether there is any other process using GPU 0.

1. If yes, please stop them, or start PaddlePaddle on another GPU.
2. If no, please decrease the batch size of your model.
   (at /paddle/paddle/fluid/memory/allocation/cuda_allocator.cc:86)

可能解决方案 / Possible solutions

附件 / Appendix

[UserWangZz]

det_limit_side_len=20000设置的太大了，显存不够

[GreatV]

长图可以试试main分支下的slice操作

PaddleOCR/doc/doc_ch/quickstart.md

Lines 256 to 294 in 6954da7

	* 使用滑动窗口进行检测和识别
	要使用滑动窗口进行光学字符识别（OCR），可以使用以下代码片段：
	```Python
	from paddleocr import PaddleOCR
	from PIL import Image, ImageDraw, ImageFont
	# 初始化OCR引擎
	ocr = PaddleOCR(use_angle_cls=True, lang="en")
	img_path = "./very_large_image.jpg"
	slice = {'horizontal_stride': 300, 'vertical_stride': 500, 'merge_x_thres': 50, 'merge_y_thres': 35}
	results = ocr.ocr(img_path, cls=True, slice=slice)
	# 加载图像
	image = Image.open(img_path).convert("RGB")
	draw = ImageDraw.Draw(image)
	font = ImageFont.truetype("./doc/fonts/simfang.ttf", size=20) # 根据需要调整大小
	# 处理并绘制结果
	for res in results:
	for line in res:
	box = [tuple(point) for point in line[0]]
	# 找出边界框
	box = [(min(point[0] for point in box), min(point[1] for point in box)),
	(max(point[0] for point in box), max(point[1] for point in box))]
	txt = line[1][0]
	draw.rectangle(box, outline="red", width=2) # 绘制矩形
	draw.text((box[0][0], box[0][1] - 25), txt, fill="blue", font=font) # 在矩形上方绘制文本
	# 保存结果
	image.save("result.jpg")
	```
	此示例初始化了启用角度分类的PaddleOCR实例，并将语言设置为英语。然后调用`ocr`方法，并使用多个参数来自定义检测和识别过程，包括处理图像切片的`slice`参数。
	要更全面地了解切片操作，请参考[切片操作文档](./slice.md)。

[kennyLSN]

长图可以试试main分支下的slice操作

PaddleOCR/doc/doc_ch/quickstart.md

Lines 256 to 294 in 6954da7

	* 使用滑动窗口进行检测和识别
	要使用滑动窗口进行光学字符识别（OCR），可以使用以下代码片段：
	```Python
	from paddleocr import PaddleOCR
	from PIL import Image, ImageDraw, ImageFont
	# 初始化OCR引擎
	ocr = PaddleOCR(use_angle_cls=True, lang="en")
	img_path = "./very_large_image.jpg"
	slice = {'horizontal_stride': 300, 'vertical_stride': 500, 'merge_x_thres': 50, 'merge_y_thres': 35}
	results = ocr.ocr(img_path, cls=True, slice=slice)
	# 加载图像
	image = Image.open(img_path).convert("RGB")
	draw = ImageDraw.Draw(image)
	font = ImageFont.truetype("./doc/fonts/simfang.ttf", size=20) # 根据需要调整大小
	# 处理并绘制结果
	for res in results:
	for line in res:
	box = [tuple(point) for point in line[0]]
	# 找出边界框
	box = [(min(point[0] for point in box), min(point[1] for point in box)),
	(max(point[0] for point in box), max(point[1] for point in box))]
	txt = line[1][0]
	draw.rectangle(box, outline="red", width=2) # 绘制矩形
	draw.text((box[0][0], box[0][1] - 25), txt, fill="blue", font=font) # 在矩形上方绘制文本
	# 保存结果
	image.save("result.jpg")
	```
	此示例初始化了启用角度分类的PaddleOCR实例，并将语言设置为英语。然后调用`ocr`方法，并使用多个参数来自定义检测和识别过程，包括处理图像切片的`slice`参数。
	要更全面地了解切片操作，请参考[切片操作文档](./slice.md)。

长图可以试试main分支下的slice操作

PaddleOCR/doc/doc_ch/quickstart.md

Lines 256 to 294 in 6954da7

	* 使用滑动窗口进行检测和识别
	要使用滑动窗口进行光学字符识别（OCR），可以使用以下代码片段：
	```Python
	from paddleocr import PaddleOCR
	from PIL import Image, ImageDraw, ImageFont
	# 初始化OCR引擎
	ocr = PaddleOCR(use_angle_cls=True, lang="en")
	img_path = "./very_large_image.jpg"
	slice = {'horizontal_stride': 300, 'vertical_stride': 500, 'merge_x_thres': 50, 'merge_y_thres': 35}
	results = ocr.ocr(img_path, cls=True, slice=slice)
	# 加载图像
	image = Image.open(img_path).convert("RGB")
	draw = ImageDraw.Draw(image)
	font = ImageFont.truetype("./doc/fonts/simfang.ttf", size=20) # 根据需要调整大小
	# 处理并绘制结果
	for res in results:
	for line in res:
	box = [tuple(point) for point in line[0]]
	# 找出边界框
	box = [(min(point[0] for point in box), min(point[1] for point in box)),
	(max(point[0] for point in box), max(point[1] for point in box))]
	txt = line[1][0]
	draw.rectangle(box, outline="red", width=2) # 绘制矩形
	draw.text((box[0][0], box[0][1] - 25), txt, fill="blue", font=font) # 在矩形上方绘制文本
	# 保存结果
	image.save("result.jpg")
	```
	此示例初始化了启用角度分类的PaddleOCR实例，并将语言设置为英语。然后调用`ocr`方法，并使用多个参数来自定义检测和识别过程，包括处理图像切片的`slice`参数。
	要更全面地了解切片操作，请参考[切片操作文档](./slice.md)。

请问这是啥问题。。
/paddleocr.py", line 728, in ocr
dt_boxes, rec_res, _ = self.call(img, cls, slice)
TypeError: call() takes from 2 to 3 positional arguments but 4 were given

[GreatV]

我又试了一下没啥问题呀

pip install git+https://github.com/PaddlePaddle/PaddleOCR.git

from paddleocr import PaddleOCR 
from PIL import Image, ImageDraw, ImageFont 

# 初始化OCR引擎 
ocr = PaddleOCR(use_angle_cls=True, lang="en") 

img_path = "./very_large_image.jpg" 
slice = {'horizontal_stride': 300, 'vertical_stride': 500, 'merge_x_thres': 50, 'merge_y_thres': 35} 
results = ocr.ocr(img_path, cls=True, slice=slice) 

# 加载图像 
image = Image.open(img_path).convert("RGB") 
draw = ImageDraw.Draw(image) 
font = ImageFont.truetype("../PaddleOCR/doc/fonts/simfang.ttf", size=20)  # 根据需要调整大小 

# 处理并绘制结果 
for res in results: 
    for line in res: 
        box = [tuple(point) for point in line[0]] 
        # 找出边界框 
        box = [(min(point[0] for point in box), min(point[1] for point in box)), 
            (max(point[0] for point in box), max(point[1] for point in box))] 
        txt = line[1][0] 
        draw.rectangle(box, outline="red", width=2)  # 绘制矩形 
        draw.text((box[0][0], box[0][1] - 25), txt, fill="blue", font=font)  # 在矩形上方绘制文本 

# 保存结果 
image.save("result.jpg")