P2PNet推理和训练

0、环境信息

Package                  Version
------------------------ ------------
certifi                  2023.11.17
charset-normalizer       3.3.2
contourpy                1.2.0
cycler                   0.12.1
easydict                 1.11
filelock                 3.13.1
fonttools                4.47.2
fsspec                   2023.12.2
idna                     3.6
importlib-resources      6.1.1
Jinja2                   3.1.3
kiwisolver               1.4.5
MarkupSafe               2.1.4
matplotlib               3.8.2
mpmath                   1.3.0
networkx                 3.2.1
numpy                    1.26.3
nvidia-cublas-cu12       12.1.3.1
nvidia-cuda-cupti-cu12   12.1.105
nvidia-cuda-nvrtc-cu12   12.1.105
nvidia-cuda-runtime-cu12 12.1.105
nvidia-cudnn-cu12        8.9.2.26
nvidia-cufft-cu12        11.0.2.54
nvidia-curand-cu12       10.3.2.106
nvidia-cusolver-cu12     11.4.5.107
nvidia-cusparse-cu12     12.1.0.106
nvidia-nccl-cu12         2.18.1
nvidia-nvjitlink-cu12    12.3.101
nvidia-nvtx-cu12         12.1.105
opencv-python            4.9.0.80
packaging                23.2
pandas                   2.2.0
pillow                   10.2.0
pip                      23.3.1
protobuf                 4.25.2
pyparsing                3.1.1
python-dateutil          2.8.2
pytz                     2023.3.post1
requests                 2.31.0
scipy                    1.12.0
setuptools               68.2.2
six                      1.16.0
sympy                    1.12
tensorboardX             2.6.2.2
torch                    2.1.2
torchvision              0.16.2
triton                   2.1.0
typing_extensions        4.9.0
tzdata                   2023.4
urllib3                  2.1.0
wheel                    0.41.2
zipp                     3.17.0

1、测试

下载模型

根据url下载，并修改模型的相关路径

运行测试

CUDA_VISIBLE_DEVICES=0 python run_test.py --weight_path ./weights/SHTechA.pth --output_dir ./logs/

报错1：

from torchvision.ops import _new_empty_tensor
connot import name '_new_empty_tensor'

原因：版本问题

解决方式：将misc.py里的if语句注释到

报错2：

img_raw = img_raw.resize((new_width, new_height), Image.ANTIALIAS)
AttributeError: module 'PIL.Image' has no attribute 'ANTIALIAS'

原因：原来是在pillow的10.0.0版本中，ANTIALIAS方法被删除了，使用新的方法即可

解决方式：

将Image.LANCZOS替换为Image.Resampling.LANCZOS

运行成功

此时需要你先创建logs文件，要不然看不到预测的图片。

2、训练

数据下载

生成list

分别修改dataset_path 和output_path 来生成两个对应的list

import os

# 设置数据集目录
dataset_path = '/dev_path/qiuzx/datasets/01CrowdCounting/part_A_final/test_data'

# 设置输出的train.txt文件路径
output_path = './test.list'

# 获取images文件夹下的所有.jpg文件
image_files = [f for f in os.listdir(os.path.join(dataset_path, 'images')) if f.endswith('.jpg')]

# 打开train.txt文件以写入模式
with open(output_path, 'w') as f:
    # 遍历每个.jpg文件并写入train.txt
    for image_file in image_files:
        # 构造图片和对应txt文件的路径
        image_path = os.path.join(dataset_path, 'images', image_file)
        txt_path = os.path.join(dataset_path, 'txt', 'GT_' + os.path.splitext(image_file)[0] + '.txt')
        
        # 检查txt文件和对应的jpg文件是否存在
        if os.path.exists(txt_path) and os.path.exists(image_path):
            # 写入train.txt文件
            f.write(f"{image_path} {txt_path}\n")
        else:
            print(f"Skipping pair: {image_path}, {txt_path} as one or both files do not exist.")

修改list的位置

运行

CUDA_VISIBLE_DEVICES=0 python train.py --data_root ../01CrowdCounting/part_A_final     --dataset_file SHHA     --epochs 3500     --lr_drop 3500     --output_dir ./logs     --checkpoints_dir ./weights/train     --tensorboard_dir ./logs     --lr 0.0001     --lr_backbone 0.00001     --batch_size 8     --eval_freq 1     --gpu_id 0

运行成功