Crowd Counting P2PNet 复现

最近用p2pnet进行了复现，还对自己做的数据集进行了训练。在复现的过程中的一些注意事项在此做了记录。
代码地址:https://github.com/TencentYoutuResearch/CrowdCounting-P2PNet
论文地址：https://arxiv.org/abs/2107.12746

P2PNET网络图

测试：
在用p2pnet 测试自己的图像数据时，需要先下载vgg的训练好的模型，因为p2pnet是由vgg提取出的特征图上进行预测，计数的。原程序中没有给vgg的模型需要自己下载。
vgg_.py

"""
import torch
import torch.nn as nn__all__ = ['VGG', 'vgg11', 'vgg11_bn', 'vgg13', 'vgg13_bn', 'vgg16', 'vgg16_bn','vgg19_bn', 'vgg19',
]model_urls = {'vgg11': 'https://download.pytorch.org/models/vgg11-bbd30ac9.pth','vgg13': 'https://download.pytorch.org/models/vgg13-c768596a.pth','vgg16': 'https://download.pytorch.org/models/vgg16-397923af.pth','vgg19': 'https://download.pytorch.org/models/vgg19-dcbb9e9d.pth','vgg11_bn': 'https://download.pytorch.org/models/vgg11_bn-6002323d.pth','vgg13_bn': 'https://download.pytorch.org/models/vgg13_bn-abd245e5.pth','vgg16_bn': 'https://download.pytorch.org/models/vgg16_bn-6c64b313.pth','vgg19_bn': 'https://download.pytorch.org/models/vgg19_bn-c79401a0.pth',
}model_paths = {'vgg16_bn': 'weights/vgg16_bn-6c64b313.pth','vgg16': '/apdcephfs/private_changanwang/checkpoints/vgg16-397923af.pth',}

这里我们可以看到在model_paths中给了vgg16_bn和vgg16两种模型，我们只需要下载其中一种模型并把路径填到里面就行了。
run_test.py

def get_args_parser():parser = argparse.ArgumentParser('Set parameters for P2PNet evaluation', add_help=False)# * Backboneparser.add_argument('--backbone', default='vgg16_bn', type=str,help="name of the convolutional backbone to use")parser.add_argument('--row', default=2, type=int,help="row number of anchor points")parser.add_argument('--line', default=2, type=int,help="line number of anchor points")parser.add_argument('--output_dir', default='vis',#添加输出位置定位的路径文件夹help='path where to save')parser.add_argument('--weight_path', default='weights/SHTechA.pth',#添加p2pnet的训练好的模型help='path where the trained weights saved')parser.add_argument('--gpu_id', default=0, type=int, help='the gpu used for evaluation')return parserdef main(args, debug=False):os.environ["CUDA_VISIBLE_DEVICES"] = '{}'.format(args.gpu_id)print(args)device = torch.device('cuda')# get the P2PNetmodel = build_model(args)# move to GPUmodel.to(device)# load trained modelif args.weight_path is not None:checkpoint = torch.load(args.weight_path, map_location='cpu')model.load_state_dict(checkpoint['model'])# convert to eval modemodel.eval()# create the pre-processing transformtransform = standard_transforms.Compose([standard_transforms.ToTensor(), standard_transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),])# set your image path hereimg_path = "./vis/demo1.jpg"#测试数据图片# load the imagesimg_raw = Image.open(img_path).convert('RGB')# round the sizewidth, height = img_raw.sizenew_width = width // 128 * 128new_height = height // 128 * 128img_raw = img_raw.resize((new_width, new_height), Image.ANTIALIAS)# pre-proccessingimg = transform(img_raw)

这里只需要添加自己的输出路径、权重路径、和测试数据的路径就行了。
注意如果出现下面问题

这是版本不统一的问题，只需要将misc.py里的if语句注释掉就行了，如下图

训练：
如果要训练自己的数据集，那需要注意，这个程序进行训练需要的是数据图像和标注位置的txt文件，注意是txt文件，不是公共数据集的mat或csv文件。

import os
import random
import torch
import numpy as np
from torch.utils.data import Dataset
from PIL import Image
import cv2
import glob
import scipy.io as ioclass SHHA(Dataset):def __init__(self, data_root, transform=None, train=False, patch=False, flip=False):self.root_path = data_rootself.train_lists = "shanghai_tech_part_a_train.list"#训练集路径self.eval_list = "shanghai_tech_part_a_test.list"#测试集路径# there may exist multiple list filesself.img_list_file = self.train_lists.split(',')if train:self.img_list_file = self.train_lists.split(',')else:self.img_list_file = self.eval_list.split(',')self.img_map = {}

这上面的训练集和测试集的文件路径，其实就将训练图片和其标注文件的路径放在一个txt文件中，最后把.txt后缀名改成.list。这个txt文件里的内容是：
图片1路径+空格+ 其对应的是标注文件txt文件的路径
图片2路径+空格+ 其对应的是标注文件txt文件的路径
…
然后改变train.py中的路径

# dataset parametersparser.add_argument('--dataset_file', default='SHHA')parser.add_argument('--data_root', default='./new_public_density_data',#你放这个程序文件夹的位置help='path where the dataset is')parser.add_argument('--output_dir', default='./log',#需要自己建立文件夹help='path where to save, empty for no saving')parser.add_argument('--checkpoints_dir', default='./ckpt',#需要自己建立文件夹，用来存放训练后的权重文件help='path where to save checkpoints, empty for no saving')parser.add_argument('--tensorboard_dir', default='./runs',help='path where to save, empty for no saving')parser.add_argument('--seed', default=42, type=int)parser.add_argument('--resume', default='', help='resume from checkpoint')parser.add_argument('--start_epoch', default=0, type=int, metavar='N',help='start epoch')parser.add_argument('--eval', action='store_true')parser.add_argument('--num_workers', default=8, type=int)parser.add_argument('--eval_freq', default=5, type=int,help='frequency of evaluation, default setting is evaluating in every 5 epoch')parser.add_argument('--gpu_id', default=0, type=int, help='the gpu used for training')return parserdef main(args):os.environ["CUDA_VISIBLE_DEVICES"] = '{}'.format(args.gpu_id)# create the logging filerun_log_name = os.path.join(args.output_dir, 'run_log.txt')#这个run_log.txt需要自己建立with open(run_log_name, "w") as log_file:log_file.write('Eval Log %s\n' % time.strftime("%c"))

改变上面的程序文件夹存放的路径，添加训练好的模型存放的文件夹路径，和最后在自己建立的output_dir文件夹中建立run_log.txt文件，就可以进行训练了。
以上就是我们在复现的过程中需要调整的信息了。
如果有哪些小伙伴出现其他问题欢迎在评论区留言。

Crowd Counting P2PNet 复现相关推荐

[CAN] [CVPR2019]：Context-Aware Crowd Counting论文+代码解读
1.论文论文链接:https://arxiv.org/pdf/1811.10452.pdf 代码链接:GitHub - weizheliu/Context-Aware-Crowd-Counting: ...
人群密度估计--Structured Inhomogeneous Density Map Learning for Crowd Counting
Structured Inhomogeneous Density Map Learning for Crowd Counting https://arxiv.org/abs/1801.06642 针对 ...
人群密度估计--Leveraging Unlabeled Data for Crowd Counting by Learning to Rank
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank CVPR2018 https://github.com/xialeil ...
人群密度估计--Crowd Counting Via Scale-adaptive Convolutional Nerual Network
Crowd Counting Via Scale-adaptive Convolutional Nerual Network https://arxiv.org/abs/1711.04433v2 Co ...
人群密度估计--Learning a perspective-embedded deconvolution network for crowd counting
Learning a perspective-embedded deconvolution network for crowd counting 没有找到代码本文在人群密度估计这个问题上的创新点: ...
越线人群计数--Crossing-line Crowd Counting with Two-phase Deep Neural Networks
Crossing-line Crowd Counting with Two-phase Deep Neural Networks ECCV2016 人群计数有两种做法:1) region-of-int ...
快速人群密度估计--Multi-scale Convolutional Neural Networks for Crowd Counting
Multi-scale Convolutional Neural Networks for Crowd Counting https://arxiv.org/abs/1702.02359 对于人群密度 ...
人群密度估计--Fully Convolutional Crowd Counting On Highly Congested Scenes
Fully Convolutional Crowd Counting On Highly Congested Scenes The 12th International Conference on C ...
人群密度估计--CrowdNet: A Deep Convolutional Network for Dense Crowd Counting
CrowdNet: A Deep Convolutional Network for Dense Crowd Counting published in the proceedings of ACM ...
人群密度估计--Spatiotemporal Modeling for Crowd Counting in Videos
Spatiotemporal Modeling for Crowd Counting in Videos ICCV2017 针对视频人群密度估计问题,这里主要侧重视频中的 temporal infor ...

Crowd Counting P2PNet 复现

Crowd Counting P2PNet 复现相关推荐

最新文章

热门文章