RocketGit

hahattpro / crowd_counting_framework (public) (License: Unspecified) (since 2020-12-23) (hash sha1)

backupfrom https://github.com/ttpro1995/crowd_counting_framework

Clone URLs: https://rocketgit.com/user/hahattpro/crowd_counting_framework ssh://rocketgit@ssh.rocketgit.com/user/hahattpro/crowd_counting_framework git://git.rocketgit.com/user/hahattpro/crowd_counting_framework

List of commits:

Subject	Hash	Author	Date (UTC)
commit all	6dad7a58f7dbf9fc288ce9dd3e92be538851c2a7	Thai Thien	2019-08-29 19:10:44
input d1,d2,d3 match	fc2a809241f8b6356d964c63d40cbebd55ca5f6c	Thai Thien	2019-08-28 17:57:05
WIP	39eab26d061e61dfffbf164dbd5fd878299b7250	thient	2019-08-28 11:09:12
output of de is ok	dd770386674df3e0fbebafdfc48a9352bc28967d	thient	2019-08-28 10:54:09
code pacnn	c49537b5cc91e96e4e35c9338d2c95b9bb41c672	Thai Thien	2019-08-27 16:35:27
crowd counting stuff	da9f27a39cba9bdd021b6b5c562f5f7c2be50190	Thai Thien	2019-08-24 18:27:44
seem ok	53fa176c31669a0e89b04adf290cb398f0316c45	Thai Thien	2019-08-24 18:26:31
flow ok	ad849681000818dfbcd0c1715c2858aed7236041	Thai Thien	2019-08-24 17:00:02
wip	23c3ec48497782bbc91d829e1c8a682502360ab9	Thai Thien	2019-08-24 14:19:22
work in progress, try to use https://pytorch.org/ignite/quickstart.html	39c824fe8fc2501628ee42c236a844df45521007	Thai Thien	2019-08-24 07:41:46
Work in progress	984be31d85e5cbdb2af296ccdb128381fe9bf09e	Thai Thien	2019-08-24 05:30:51
README	045706df1fa3452d150a190675c60e80ebd18e08	Thai Thien	2019-08-24 04:11:49
init	18f7c296fb05c4340c1c20ca84c60fef4f93bb1f	Thai Thien	2019-08-24 04:08:16

Commit 6dad7a58f7dbf9fc288ce9dd3e92be538851c2a7 - commit all
Author: Thai Thien
Author date (UTC): 2019-08-29 19:10
Committer name: Thai Thien
Committer date (UTC): 2019-08-29 19:10
Parent(s): fc2a809241f8b6356d964c63d40cbebd55ca5f6c
Signer:
Signing key: 2F827CFB43A19EA3
Signing status: E
Tree: 9fa8fe6f03626db02f09f2eb111334c9215dc839

File	Lines added	Lines deleted
args_util.py	2	2
data_flow.py	19	2
main_pacnn_ucfcc50.py	29	15
models/pacnn.py	1	1
pytorch_ssim/__init__.py	73	0

File args_util.py changed (mode: 100644) (index f5fd17d..6728e74)
...	...	def real_args_parse():
67	67	parser.add_argument('--model', action="store", default="csrnet")	parser.add_argument('--model', action="store", default="csrnet")
68	68
69	69	# args with default value	# args with default value
70		parser.add_argument('--lr', action="store", default=1e-7, type=float)
	70		parser.add_argument('--lr', action="store", default=1e-8, type=float)
71	71	parser.add_argument('--momentum', action="store", default=0.95, type=float)	parser.add_argument('--momentum', action="store", default=0.95, type=float)
72		parser.add_argument('--decay', action="store", default=5*1e-4, type=float)
	72		parser.add_argument('--decay', action="store", default=5*1e-3, type=float)
73	73	parser.add_argument('--epochs', action="store", default=1, type=int)	parser.add_argument('--epochs', action="store", default=1, type=int)
74	74
75	75	# args.original_lr = 1e-7	# args.original_lr = 1e-7

File data_flow.py changed (mode: 100644) (index e9a9443..2d70582)
...	...	def load_data_ucf_cc50_pancnn(img_path, train=True):
74	74	gt_file = h5py.File(gt_path, 'r')	gt_file = h5py.File(gt_path, 'r')
75	75	target = np.asarray(gt_file['density'])	target = np.asarray(gt_file['density'])
76	76
	77		if train:
	78		crop_size = (int(img.size[0] / 2), int(img.size[1] / 2))
	79		if random.randint(0, 9) <= -1:
	80
	81		dx = int(random.randint(0, 1) * img.size[0] * 1. / 2)
	82		dy = int(random.randint(0, 1) * img.size[1] * 1. / 2)
	83		else:
	84		dx = int(random.random() * img.size[0] * 1. / 2)
	85		dy = int(random.random() * img.size[1] * 1. / 2)
	86
	87		img = img.crop((dx, dy, crop_size[0] + dx, crop_size[1] + dy))
	88		target = target[dy:crop_size[1] + dy, dx:crop_size[0] + dx]
	89
	90		if random.random() > 0.8:
	91		target = np.fliplr(target)
	92		img = img.transpose(Image.FLIP_LEFT_RIGHT)
	93
77	94	target1 = cv2.resize(target, (int(target.shape[1] / 8), int(target.shape[0] / 8)),	target1 = cv2.resize(target, (int(target.shape[1] / 8), int(target.shape[0] / 8)),
78	95	interpolation=cv2.INTER_CUBIC) * 64	interpolation=cv2.INTER_CUBIC) * 64
79	96	target2 = cv2.resize(target, (int(target.shape[1] / 16), int(target.shape[0] / 16)),	target2 = cv2.resize(target, (int(target.shape[1] / 16), int(target.shape[0] / 16)),
80		interpolation=cv2.INTER_CUBIC) * 64*2
	97		interpolation=cv2.INTER_CUBIC) * 64 #*2
81	98	target3 = cv2.resize(target, (int(target.shape[1] / 32), int(target.shape[0] / 32)),	target3 = cv2.resize(target, (int(target.shape[1] / 32), int(target.shape[0] / 32)),
82		interpolation=cv2.INTER_CUBIC) * 64*4
	99		interpolation=cv2.INTER_CUBIC) * 64 #*4
83	100
84	101	return img, (target1, target2, target3)	return img, (target1, target2, target3)
85	102

File main_pacnn_ucfcc50.py changed (mode: 100644) (index ab19946..f125b41)
...	...	import os
11	11	import cv2	import cv2
12	12	from torchvision import datasets, transforms	from torchvision import datasets, transforms
13	13	from data_flow import ListDataset	from data_flow import ListDataset
	14		import pytorch_ssim
14	15
15	16	if __name__ == "__main__":	if __name__ == "__main__":
16		# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
17		device = "cpu"
	17		device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
	18		# device = "cpu"
18	19	print(device)	print(device)
19	20	args = real_args_parse()	args = real_args_parse()
20	21	print(args)	print(args)

...	...	if __name__ == "__main__":
40	41	batch_size=1, num_workers=4)	batch_size=1, num_workers=4)
41	42
42	43	# create model	# create model
43		net = PACNN()
44
45		for train_img, label in train_loader_pacnn:
46		d1_label, d2_label, d3_label = label
47		d1, d2, d3 = net(train_img)
48		print(d1.size())
49		print(d2.size())
50		print(d3.size())
51		print("====")
52		print(d1_label.size())
53		print(d2_label.size())
54		print(d3_label.size())
55		print("done===done=====")
	44		net = PACNN().to(device)
	45		criterion_mse = nn.MSELoss(size_average=False).to(device)
	46		criterion_ssim = pytorch_ssim.SSIM(window_size=11).to(device)
	47
	48
	49		optimizer = torch.optim.SGD(net.parameters(), args.lr,
	50		momentum=args.momentum,
	51		weight_decay=args.decay)
	52		for e in range(3):
	53		loss_sum = 0
	54		for train_img, label in train_loader_pacnn:
	55		d1_label, d2_label, d3_label = label
	56		d1_label = d1_label.to(device)
	57		d2_label = d2_label.to(device)
	58		d3_label = d3_label.to(device)
	59		d1, d2, d3 = net(train_img.to(device))
	60		loss_1 = criterion_mse(d1, d1_label) + criterion_ssim(d1.unsqueeze(0), d1_label.unsqueeze(0))
	61		loss_2 = criterion_mse(d2, d2_label) + criterion_ssim(d2.unsqueeze(0), d2_label.unsqueeze(0))
	62		loss_3 = criterion_mse(d3, d3_label) + criterion_ssim(d3.unsqueeze(0), d3_label.unsqueeze(0))
	63
	64		loss = loss_1 + loss_2 + loss_3
	65		loss.backward()
	66		optimizer.step()
	67		loss_sum += loss.item()
	68		avg_loss = loss_sum/40
	69		print(avg_loss)

File models/pacnn.py changed (mode: 100644) (index a8b3759..7144f8d)
...	...	class PACNN(nn.Module):
25	25	de1 = self.de1_11((self.de1net(x)))	de1 = self.de1_11((self.de1net(x)))
26	26	de2 = self.de2_11((self.de2net(x)))	de2 = self.de2_11((self.de2net(x)))
27	27	de3 = self.de3_11((self.de3net(x)))	de3 = self.de3_11((self.de3net(x)))
28		return de1, de2, de3
	28		return de1.squeeze(0), de2.squeeze(0), de3.squeeze(0)
29	29
30	30
31	31	if __name__ == "__main__":	if __name__ == "__main__":

File pytorch_ssim/__init__.py added (mode: 100644) (index 0000000..738e803)
	1	import torch
	2	import torch.nn.functional as F
	3	from torch.autograd import Variable
	4	import numpy as np
	5	from math import exp
	6
	7	def gaussian(window_size, sigma):
	8	gauss = torch.Tensor([exp(-(x - window_size//2)*2/float(2sigma**2)) for x in range(window_size)])
	9	return gauss/gauss.sum()
	10
	11	def create_window(window_size, channel):
	12	_1D_window = gaussian(window_size, 1.5).unsqueeze(1)
	13	_2D_window = _1D_window.mm(_1D_window.t()).float().unsqueeze(0).unsqueeze(0)
	14	window = Variable(_2D_window.expand(channel, 1, window_size, window_size).contiguous())
	15	return window
	16
	17	def _ssim(img1, img2, window, window_size, channel, size_average = True):
	18	mu1 = F.conv2d(img1, window, padding = window_size//2, groups = channel)
	19	mu2 = F.conv2d(img2, window, padding = window_size//2, groups = channel)
	20
	21	mu1_sq = mu1.pow(2)
	22	mu2_sq = mu2.pow(2)
	23	mu1_mu2 = mu1*mu2
	24
	25	sigma1_sq = F.conv2d(img1*img1, window, padding = window_size//2, groups = channel) - mu1_sq
	26	sigma2_sq = F.conv2d(img2*img2, window, padding = window_size//2, groups = channel) - mu2_sq
	27	sigma12 = F.conv2d(img1*img2, window, padding = window_size//2, groups = channel) - mu1_mu2
	28
	29	C1 = 0.01**2
	30	C2 = 0.03**2
	31
	32	ssim_map = ((2mu1_mu2 + C1)(2sigma12 + C2))/((mu1_sq + mu2_sq + C1)(sigma1_sq + sigma2_sq + C2))
	33
	34	if size_average:
	35	return ssim_map.mean()
	36	else:
	37	return ssim_map.mean(1).mean(1).mean(1)
	38
	39	class SSIM(torch.nn.Module):
	40	def __init__(self, window_size = 11, size_average = True):
	41	super(SSIM, self).__init__()
	42	self.window_size = window_size
	43	self.size_average = size_average
	44	self.channel = 1
	45	self.window = create_window(window_size, self.channel)
	46
	47	def forward(self, img1, img2):
	48	(_, channel, _, _) = img1.size()
	49
	50	if channel == self.channel and self.window.data.type() == img1.data.type():
	51	window = self.window
	52	else:
	53	window = create_window(self.window_size, channel)
	54
	55	if img1.is_cuda:
	56	window = window.cuda(img1.get_device())
	57	window = window.type_as(img1)
	58
	59	self.window = window
	60	self.channel = channel
	61
	62
	63	return _ssim(img1, img2, window, self.window_size, channel, self.size_average)
	64
	65	def ssim(img1, img2, window_size = 11, size_average = True):
	66	(_, channel, _, _) = img1.size()
	67	window = create_window(window_size, channel)
	68
	69	if img1.is_cuda:
	70	window = window.cuda(img1.get_device())
	71	window = window.type_as(img1)
	72
	73	return _ssim(img1, img2, window, window_size, channel, size_average)

Hints:
Before first commit, do not forget to setup your git environment:

git config --global user.name "your_name_here"
git config --global user.email "your@email_here"

Clone this repository using HTTP(S):

git clone https://rocketgit.com/user/hahattpro/crowd_counting_framework

Clone this repository using ssh (do not forget to upload a key first):

git clone ssh://rocketgit@ssh.rocketgit.com/user/hahattpro/crowd_counting_framework

Clone this repository using git:

git clone git://git.rocketgit.com/user/hahattpro/crowd_counting_framework

You are allowed to anonymously push to this repository.
This means that your pushed commits will automatically be transformed into a merge request:

... clone the repository ...
... make some changes and some commits ...
git push origin main