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)
m3 t123 exp	f76cec015c34648ec45f02865562f6d05d507eb2	Thai Thien	2020-04-03 15:27:22
m1 shb	35455b092280af7e7f3140601b909dc58b5c2590	Thai Thien	2020-04-02 10:42:43
m2	765cc07c3d3645b897f53b54f905b8a9c3aa0ac3	Thai Thien	2020-04-02 10:37:47
typo again	3e12138f4effbbb0fb8dc1e5aa98c7f8a689d737	Thai Thien	2020-04-02 10:34:01
ok dude	19a59daa4433b0dc4fb24023dff428928090ab7b	Thai Thien	2020-04-02 10:31:17
typo	91cbb50105fb85140e9e34261bbaf580f788425d	Thai Thien	2020-04-02 10:29:55
1 debug trainning script	5f075f3774502190ecb71f663222ee0808b5a2ef	Thai Thien	2020-04-02 10:28:05
setting up code	6e83edd7283bfeec6ec1a6e5057719e8ae95e8ea	Thai Thien	2020-04-02 10:25:18
fix typo	f7565a25dc7058d1f344cc0d6e610fa2bf8f19e9	Thai Thien	2020-04-01 10:58:38
t4 aug	36b5bbc17111dca87cfa082f5a512b84576c8fca	Thai Thien	2020-04-01 10:55:12
reproduce ccnn without batchnorm	e4678b328ba9d00b9d94617a1cd0f34732f7b2ac	Thai Thien	2020-03-30 16:48:35
change gpu number	f4c616536d13d4caf383166caae0d8298cb73738	Thai Thien	2020-03-28 04:17:37
train custom ccnn_v2	8b6fca9cd3cb15111befd7332bc7e764d5494a66	Thai Thien	2020-03-28 03:51:07
v5	24e4a8a9c8948c21b0d6beadec0418adceec8577	Thai Thien	2020-03-20 17:56:32
hope it fix cuda bug	7b8cbb70550df279035ef56a22450d17fed6b29a	Thai Thien	2020-03-20 17:55:24
ccnnv4 sha	a4acad0caf5a18fc9ea037a5e3d6db9d6b6106c7	Thai Thien	2020-03-20 17:38:30
change gpu	fa2c1c54858d8a57585c3801dc168c464ba689dd	Thai Thien	2020-03-18 17:31:38
fix typo again	d71ce17c4efbcc2cd31917f0e27e90bd964ac964	Thai Thien	2020-03-18 17:30:34
typo again	c16871fb2cc035e168d62df9c5d4401bc465df95	Thai Thien	2020-03-18 17:27:45
fix minor typo	01d907dda5c1e4c08b5ca5607aa77365cfd70a2e	Thai Thien	2020-03-18 17:25:32

Commit f76cec015c34648ec45f02865562f6d05d507eb2 - m3 t123 exp
Author: Thai Thien
Author date (UTC): 2020-04-03 15:27
Committer name: Thai Thien
Committer date (UTC): 2020-04-03 15:27
Parent(s): 35455b092280af7e7f3140601b909dc58b5c2590
Signer:
Signing key:
Signing status: N
Tree: 88da5a17b37cdc50ff844a9a8bc0bb7b9f66eb91

File	Lines added	Lines deleted
experiment_meow_main.py	5	1
models/meow_experiment/kitten_meow_1.py	69	0
train_script/meow_one/m2_t1_sha.sh	0	0
train_script/meow_one/m3_t1_sha.sh	3	3
train_script/meow_one/m3_t2_sha.sh	4	4
train_script/meow_one/m3_t3_sha.sh	13	0

File experiment_meow_main.py changed (mode: 100644) (index 7b76dc2..3852e31)
...	...	from visualize_util import get_readable_time
11	11
12	12	import torch	import torch
13	13	from torch import nn	from torch import nn
14		from models.meow_experiment.kitten_meow_1 import M1, M2
	14		from models.meow_experiment.kitten_meow_1 import M1, M2, M3
15	15	import os	import os
16	16	from model_util import get_lr	from model_util import get_lr
17	17
18	18	COMET_ML_API = "S3mM1eMq6NumMxk2QJAXASkUM"	COMET_ML_API = "S3mM1eMq6NumMxk2QJAXASkUM"
19	19	PROJECT_NAME = "meow-one-experiment-insita"	PROJECT_NAME = "meow-one-experiment-insita"
20	20
	21
21	22	def very_simple_param_count(model):	def very_simple_param_count(model):
22	23	result = sum([p.numel() for p in model.parameters()])	result = sum([p.numel() for p in model.parameters()])
23	24	return result	return result
24	25
	26
25	27	if __name__ == "__main__":	if __name__ == "__main__":
26	28	experiment = Experiment(project_name=PROJECT_NAME, api_key=COMET_ML_API)	experiment = Experiment(project_name=PROJECT_NAME, api_key=COMET_ML_API)
27	29	device = torch.device("cuda" if torch.cuda.is_available() else "cpu")	device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

...	...	if __name__ == "__main__":
61	63	model = M1()	model = M1()
62	64	elif model_name == "M2":	elif model_name == "M2":
63	65	model = M2()	model = M2()
	66		elif model_name == "M3":
	67		model = M3()
64	68	else:	else:
65	69	print("error: you didn't pick a model")	print("error: you didn't pick a model")
66	70	exit(-1)	exit(-1)

File models/meow_experiment/kitten_meow_1.py changed (mode: 100644) (index 055ce5f..43169dd)
...	...	from models.deform_conv_v2 import DeformConv2d, TorchVisionBasicDeformConv2d
5	5	import collections	import collections
6	6	import torch.nn.functional as F	import torch.nn.functional as F
7	7
	8		class M3(nn.Module):
	9		"""
	10		A REAL-TIME DEEP NETWORK FOR CROWD COUNTING
	11		https://arxiv.org/pdf/2002.06515.pdf
	12		the improve version
	13
	14		we change 5x5 7x7 9x9 with 3x3
	15		Keep the tail
	16		"""
	17		def __init__(self, load_weights=False):
	18		super(M3, self).__init__()
	19		self.model_note = "We replace 5x5 7x7 9x9 with 3x3, no batchnorm yet, tail dilated"
	20		# self.red_cnn = nn.Conv2d(3, 10, 9, padding=4)
	21		# self.green_cnn = nn.Conv2d(3, 14, 7, padding=3)
	22		# self.blue_cnn = nn.Conv2d(3, 16, 5, padding=2)
	23
	24		# ideal from crowd counting using DMCNN
	25		self.front_cnn_1 = nn.Conv2d(3, 20, 3, padding=1)
	26		self.front_cnn_2 = nn.Conv2d(20, 16, 3, padding=1)
	27		self.front_cnn_3 = nn.Conv2d(16, 14, 3, padding=1)
	28		self.front_cnn_4 = nn.Conv2d(14, 10, 3, padding=1)
	29
	30		self.c0 = nn.Conv2d(40, 40, 3, padding=1)
	31		self.max_pooling = nn.MaxPool2d(kernel_size=2, stride=2)
	32
	33		self.c1 = nn.Conv2d(40, 60, 3, padding=1)
	34
	35		# ideal from CSRNet
	36		self.c2 = nn.Conv2d(60, 40, 3, padding=2, dilation=2)
	37		self.c3 = nn.Conv2d(40, 20, 3, padding=2, dilation=2)
	38		self.c4 = nn.Conv2d(20, 10, 3, padding=2, dilation=2)
	39		self.output = nn.Conv2d(10, 1, 1)
	40
	41		def forward(self,x):
	42		#x_red = self.max_pooling(F.relu(self.red_cnn(x), inplace=True))
	43		#x_green = self.max_pooling(F.relu(self.green_cnn(x), inplace=True))
	44		#x_blue = self.max_pooling(F.relu(self.blue_cnn(x), inplace=True))
	45
	46		x_red = F.relu(self.front_cnn_1(x), inplace=True)
	47		x_red = F.relu(self.front_cnn_2(x_red), inplace=True)
	48		x_red = F.relu(self.front_cnn_3(x_red), inplace=True)
	49		x_red = F.relu(self.front_cnn_4(x_red), inplace=True)
	50		x_red = self.max_pooling(x_red)
	51
	52		x_green = F.relu(self.front_cnn_1(x), inplace=True)
	53		x_green = F.relu(self.front_cnn_2(x_green), inplace=True)
	54		x_green = F.relu(self.front_cnn_3(x_green), inplace=True)
	55		x_green = self.max_pooling(x_green)
	56
	57		x_blue = F.relu(self.front_cnn_1(x), inplace=True)
	58		x_blue = F.relu(self.front_cnn_2(x_blue), inplace=True)
	59		x_blue = self.max_pooling(x_blue)
	60
	61		x = torch.cat((x_red, x_green, x_blue), 1)
	62		x = F.relu(self.c0(x), inplace=True)
	63
	64		x = F.relu(self.c1(x), inplace=True)
	65
	66		x = F.relu(self.c2(x), inplace=True)
	67		x = self.max_pooling(x)
	68
	69		x = F.relu(self.c3(x), inplace=True)
	70		x = self.max_pooling(x)
	71
	72		x = F.relu(self.c4(x), inplace=True)
	73
	74		x = self.output(x)
	75		return x
	76
8	77
9	78	class M1(nn.Module):	class M1(nn.Module):
10	79	"""	"""

File train_script/meow_one/m2_t1_sha.sh copied from file train_script/meow_one/m1_t2_sha.sh (similarity 100%)

File train_script/meow_one/m3_t1_sha.sh copied from file train_script/meow_one/m1_t1_sha.sh (similarity 72%) (mode: 100644) (index bcdb158..a01a2dd)
1		task="m1_t1_sha"
	1		task="m3_t1_sha"
2	2
3	3	CUDA_VISIBLE_DEVICES=1 HTTPS_PROXY="http://10.30.58.36:81" nohup python experiment_meow_main.py \	CUDA_VISIBLE_DEVICES=1 HTTPS_PROXY="http://10.30.58.36:81" nohup python experiment_meow_main.py \
4	4	--task_id $task \	--task_id $task \
5		--note "m1 sha shanghaitech_keepfull with decay, change the head, keep the tail " \
6		--model "M1" \
	5		--note "m3 sha shanghaitech_keepfull with decay, change the head, tail dilated" \
	6		--model "M3" \
7	7	--input /data/rnd/thient/thient_data/ShanghaiTech/part_A \	--input /data/rnd/thient/thient_data/ShanghaiTech/part_A \
8	8	--lr 1e-4 \	--lr 1e-4 \
9	9	--decay 1e-4 \	--decay 1e-4 \

File train_script/meow_one/m3_t2_sha.sh renamed from train_script/meow_one/m1_t2_sha.sh (similarity 67%) (mode: 100644) (index a798494..e64545a)
1		task="m2_t1_sha"
	1		task="m3_t2_sha"
2	2
3	3	CUDA_VISIBLE_DEVICES=2 HTTPS_PROXY="http://10.30.58.36:81" nohup python experiment_meow_main.py \	CUDA_VISIBLE_DEVICES=2 HTTPS_PROXY="http://10.30.58.36:81" nohup python experiment_meow_main.py \
4	4	--task_id $task \	--task_id $task \
5		--note "m2 sha shanghaitech_keepfull with decay, keep the head but change the tail " \
6		--model "M2" \
	5		--note "m3 sha shanghaitech_keepfull with decay, change the head, tail dilated. x10 decay " \
	6		--model "M3" \
7	7	--input /data/rnd/thient/thient_data/ShanghaiTech/part_A \	--input /data/rnd/thient/thient_data/ShanghaiTech/part_A \
8	8	--lr 1e-4 \	--lr 1e-4 \
9		--decay 1e-4 \
	9		--decay 1e-3 \
10	10	--datasetname shanghaitech_keepfull \	--datasetname shanghaitech_keepfull \
11	11	--epochs 502 > logs/$task.log &	--epochs 502 > logs/$task.log &
12	12

File train_script/meow_one/m3_t3_sha.sh added (mode: 100644) (index 0000000..0131e28)
	1	task="m3_t3_sha"
	2
	3	CUDA_VISIBLE_DEVICES=3 HTTPS_PROXY="http://10.30.58.36:81" nohup python experiment_meow_main.py \
	4	--task_id $task \
	5	--note "m3 sha shanghaitech_keepfull with decay, change the head, tail dilated. x10 decay and x10 lr " \
	6	--model "M3" \
	7	--input /data/rnd/thient/thient_data/ShanghaiTech/part_A \
	8	--lr 1e-3 \
	9	--decay 1e-3 \
	10	--datasetname shanghaitech_keepfull \
	11	--epochs 502 > logs/$task.log &
	12
	13	echo logs/$task.log # for convenience

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