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)
flatten collate	c04708ae0defc81dbf441395e1d27de6a1d598fc	Thai Thien	2020-07-12 13:54:01
travis remove nightly	d4b0c714823046bbafcd3c816d56f7079c76d126	Thai Thien	2020-07-12 13:27:55
travis	e6368ec3102e01f1bdc71a80a78f0db3617d7e08	Thai Thien	2020-07-12 12:33:21
flatten_collate	1e460396875c205c42de27449f56e73cd4ec10e0	Thai Thien	2020-07-12 12:23:40
train val split test ratio to 0.1	5091da3f0b45d875a38c2829e4fec5e61116e869	Thai Thien	2020-07-11 03:14:26
f	1defee5dc452c2da5fb540ff1050f6e01fe1878b	Thai Thien	2020-07-10 17:04:58
fix log	db4d655a313b8f3951baf225a4b197fce4bcdd4b	Thai Thien	2020-07-10 16:50:03
typo, trunc 3.0 not 4.0	ee29c49efd0a5087f11662997b8992d25671a33a	Thai Thien	2020-07-10 16:44:55
let try with trunc 3.0	f8eac179b1fa79d6451f19a9e6a35b82b94646a4	Thai Thien	2020-07-10 16:43:41
train script	7e6984a9eaa6609182ad6c786f6742dd29f1d017	Thai Thien	2020-06-14 10:23:05
fix H3	71d41ac82f273857f7643d8e63e502c701888a5b	Thai Thien	2020-06-14 10:10:39
typo	bc78059e62b17a5a617e3e0294f5c72dd4bd347e	Thai Thien	2020-06-14 10:06:35
H3_t1_sha H3_t2_sha H3_t2_shb_fixed ccnn_adam_t7_shb ccnn_adam_t7_sha	ca7e1ccab41314f8a30e48609421db3a2c41bbe0	Thai Thien	2020-06-14 09:59:17
train script	da9ad066d6d835f38eb9e9c2b771ff412124552a	Thai Thien	2020-06-13 18:06:22
train script ccnnv9	8ab76b70dad11b1458f8c1bfc2f1765c6ece984d	Thai Thien	2020-06-13 17:57:02
train script	7ab6cf827c811af9a0c9168de2fde4dbd0fb3e61	Thai Thien	2020-06-13 11:54:21
fix typo also	3d32c83c822a66ac24ebb6d8cfe4d023d167d441	Thai Thien	2020-06-13 04:45:27
typo in training script	07c6055e3bdacaca34c0aaacd49d5c87a07f4320	Thai Thien	2020-06-13 04:44:20
ccnnv9 leaky relu	5fd462122ef62498a51dd83db2c89507254d7a96	Thai Thien	2020-06-13 04:40:41
CompactCNNv9 leaky relu	03e5f342cc6eb3ba4fd0b98061398acedde6ab54	Thai Thien	2020-06-13 04:36:43

Commit c04708ae0defc81dbf441395e1d27de6a1d598fc - flatten collate
Author: Thai Thien
Author date (UTC): 2020-07-12 13:54
Committer name: Thai Thien
Committer date (UTC): 2020-07-12 13:54
Parent(s): d4b0c714823046bbafcd3c816d56f7079c76d126
Signing key:
Tree: f2601ad6d1ff47a8d60e2d247d3c1cee9f4afdab

File	Lines added	Lines deleted
data_flow.py	60	3

File data_flow.py changed (mode: 100644) (index 0ea457e..07a3340)
...	...	from PIL import Image
18	18	import torchvision.transforms.functional as F	import torchvision.transforms.functional as F
19	19	from torchvision import datasets, transforms	from torchvision import datasets, transforms
20	20	import scipy.io # import scipy does not work https://stackoverflow.com/questions/11172623/import-problems-with-scipy-io	import scipy.io # import scipy does not work https://stackoverflow.com/questions/11172623/import-problems-with-scipy-io
21		from data_util.dataset_utils import my_collate
	21		from data_util.dataset_utils import my_collate, flatten_collate
22	22
23	23	"""	"""
24	24	create a list of file (full directory)	create a list of file (full directory)

...	...	def load_data_shanghaitech_60p_random(img_path, train=True):
430	430
431	431	return img, target1	return img, target1
432	432
	433
	434		def load_data_shanghaitech_non_overlap(img_path, train=True):
	435		"""
	436		per sample, crop 4, non-overlap
	437		:param img_path:
	438		:param train:
	439		:return:
	440		"""
	441		gt_path = img_path.replace('.jpg', '.h5').replace('images', 'ground-truth-h5')
	442		img = Image.open(img_path).convert('RGB')
	443		gt_file = h5py.File(gt_path, 'r')
	444		target = np.asarray(gt_file['density'])
	445		target_factor = 8
	446
	447		if train:
	448		# for each image
	449		# create 8 patches, 4 non-overlap 4 corner
	450		# for each of 4 patch, create another 4 flip
	451		crop_img = []
	452		crop_label = []
	453		for i in range(2):
	454		for j in range(2):
	455		crop_size = (int(img.size[0] / 2), int(img.size[1] / 2))
	456
	457		# crop non-overlap
	458		dx = int(i * img.size[0] * 1. / 2)
	459		dy = int(j * img.size[1] * 1. / 2)
	460		img = img.crop((dx, dy, crop_size[0] + dx, crop_size[1] + dy))
	461		target = target[dy:crop_size[1] + dy, dx:crop_size[0] + dx]
	462
	463		# flip
	464		for x in range(2):
	465		if x==1:
	466		target = np.fliplr(target)
	467		img = img.transpose(Image.FLIP_LEFT_RIGHT)
	468
	469
	470		target1 = cv2.resize(target, (int(target.shape[1] / target_factor), int(target.shape[0] / target_factor)),
	471		interpolation=cv2.INTER_CUBIC) * target_factor * target_factor
	472		# target1 = target1.unsqueeze(0) # make dim (batch size, channel size, x, y) to make model output
	473		target1 = np.expand_dims(target1, axis=0) # make dim (batch size, channel size, x, y) to make model output
	474		crop_img.append(img)
	475		crop_label.append(target1)
	476		return crop_img, crop_label
	477
	478		if not train:
	479		# get correct people head count from head annotation
	480		mat_path = img_path.replace('.jpg', '.mat').replace('images', 'ground-truth').replace('IMG', 'GT_IMG')
	481		gt_count = count_gt_annotation_sha(mat_path)
	482		return img, gt_count
	483
	484
433	485	def load_data_shanghaitech_crop_random(img_path, train=True):	def load_data_shanghaitech_crop_random(img_path, train=True):
434	486	"""	"""
435	487	40 percent crop	40 percent crop

...	...	def load_data_shanghaitech_same_size_density_map(img_path, train=True):
568	620	target1 = np.expand_dims(target1, axis=0) # make dim (batch size, channel size, x, y) to make model output	target1 = np.expand_dims(target1, axis=0) # make dim (batch size, channel size, x, y) to make model output
569	621	return img, target1	return img, target1
570	622
	623
571	624	def load_data_shanghaitech_keepfull(img_path, train=True):	def load_data_shanghaitech_keepfull(img_path, train=True):
572	625	gt_path = img_path.replace('.jpg', '.h5').replace('images', 'ground-truth-h5')	gt_path = img_path.replace('.jpg', '.h5').replace('images', 'ground-truth-h5')
573	626	img = Image.open(img_path).convert('RGB')	img = Image.open(img_path).convert('RGB')

...	...	class ListDataset(Dataset):
857	910	self.load_data_fn = load_data_shanghaitech_180	self.load_data_fn = load_data_shanghaitech_180
858	911	elif dataset_name == "shanghaitech_256":	elif dataset_name == "shanghaitech_256":
859	912	self.load_data_fn = load_data_shanghaitech_256	self.load_data_fn = load_data_shanghaitech_256
	913		elif dataset_name == "shanghaitech_non_overlap":
	914		self.load_data_fn = load_data_shanghaitech_non_overlap
860	915	elif dataset_name == "ucf_cc_50":	elif dataset_name == "ucf_cc_50":
861	916	self.load_data_fn = load_data_ucf_cc50	self.load_data_fn = load_data_ucf_cc50
862	917	elif dataset_name == "ucf_cc_50_pacnn":	elif dataset_name == "ucf_cc_50_pacnn":

...	...	def get_dataloader(train_list, val_list, test_list, dataset_name="shanghaitech",
894	949	transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406],	transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406],
895	950	std=[0.229, 0.224, 0.225]),	std=[0.229, 0.224, 0.225]),
896	951	])	])
897
	952		train_collate_fn = my_collate
	953		if dataset_name == "shanghaitech_non_overlap":
	954		train_collate_fn = flatten_collate
898	955	train_loader = torch.utils.data.DataLoader(	train_loader = torch.utils.data.DataLoader(
899	956	ListDataset(train_list,	ListDataset(train_list,
900	957	shuffle=True,	shuffle=True,

...	...	def get_dataloader(train_list, val_list, test_list, dataset_name="shanghaitech",
905	962	dataset_name=dataset_name),	dataset_name=dataset_name),
906	963	batch_size=batch_size,	batch_size=batch_size,
907	964	num_workers=0,	num_workers=0,
908		collate_fn=my_collate, pin_memory=False)
	965		collate_fn=train_collate_fn, pin_memory=False)
909	966
910	967	train_loader_for_eval = torch.utils.data.DataLoader(	train_loader_for_eval = torch.utils.data.DataLoader(
911	968	ListDataset(train_list,	ListDataset(train_list,

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