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)
decide net param count	5aeadf6ad9047edded67fffd0b6584d2e9eeb777	Thai Thien	2020-12-23 16:30:14
make video code	d0737d7abec4abc900dc7e7981ca46e61f0d4153	Thai Thien	2020-12-23 16:30:04
create success	99951e32fedfcdb2ebc272d1678c094998c499f8	Thai Thien	2020-12-18 17:17:34
bike_video_frame_404_dccnn_t3	bbbd4ad319ac87586afc6ce6675adcf9d5f41fef	Thai Thien	2020-12-18 15:53:26
fix load model stuff	8c5c9c985fb747dc6392c86aa5f49282f8d12297	Thai Thien	2020-12-18 15:44:21
import comet first	dd4426221a95d0916c8cc7c02c7587ada1d910b5	Thai Thien	2020-12-18 15:41:59
predict on frame from video	ec39a97372a15ad75811fdb3417d2e48078a30b2	Thai Thien	2020-12-18 15:40:36
fix normalize in data loader	56a64c14dc62c96aba532d233ff2782414738d05	Thai Thien	2020-12-17 15:36:22
change gpu 5	46d32b3fc05909b61f56389b01c2e50e52937d9b	Thai Thien	2020-12-16 16:40:47
eval_dccnn_bike_video_VID_20201204_133931_404_model_adamw1_bigtail13i_t4_bike20	95bb636a0b544b7a72fc4c65acb47472da468766	Thai Thien	2020-12-16 16:35:14
eval_dccnn_bike_video_VID_20201204_133931_404_model_adamw1_bigtail13i_t3_bike20	27740a0e21ca42490f5cfe516b2453f7be3a2483	Thai Thien	2020-12-16 16:17:12
batch 2	85433ffd51070572d7e743a2b4ef53f340b658d4	Thai Thien	2020-12-15 16:57:56
reduce batch size 3	d2f145009c06f5e6e6e1e0270878ae275b611c2f	Thai Thien	2020-12-15 16:55:46
t3 and t4	70e09414713ea22800d8e87a0dd984f18406c375	Thai Thien	2020-12-15 16:50:32
prepare bike 20s q100	556a4d8f695ca154a5db42b363a276d7ab233434	Thai Thien	2020-12-15 16:46:08
VID_20201204_133931_404	8264dcb4d5389ed8ef2920eb3419cfc505b6a09c	Thai Thien	2020-12-13 11:31:10
add VID_20201204_133931_404	a36ba35ca7c54dcf552d17a08fa8c87720ce1c69	Thai Thien	2020-12-13 11:30:24
print total length on console	6a956d93489926e99909d234bf4a7b13d39aad56	Thai Thien	2020-12-13 11:24:56
fix input of save_density_map	dc2f014fdd570909b811ac77629f0c8c0ca156da	Thai Thien	2020-12-13 11:21:37
detach cpu for save_density_map	f41362501a0bc3578de0a48e6cbe97966a66677c	Thai Thien	2020-12-13 11:19:45

Commit 5aeadf6ad9047edded67fffd0b6584d2e9eeb777 - decide net param count
Author: Thai Thien
Author date (UTC): 2020-12-23 16:30
Committer name: Thai Thien
Committer date (UTC): 2020-12-23 16:30
Parent(s): d0737d7abec4abc900dc7e7981ca46e61f0d4153
Signing key:
Tree: f9752918cc07b36a90aeba8856bbf7697d8baba6

File	Lines added	Lines deleted
models/decide_net.py	85	0

File models/decide_net.py added (mode: 100644) (index 0000000..1040c6b)
	1	import torch.nn as nn
	2	import torch
	3	from torchvision import models
	4	from torchvision.models.detection import FasterRCNN
	5	import torchvision
	6	from torchvision.models.detection.rpn import AnchorGenerator
	7
	8	def very_simple_param_count(model):
	9	result = sum([p.numel() for p in model.parameters()])
	10	return result
	11
	12	class DecideNet(nn.Module):
	13	"""
	14	Purpose this model is to count number of parameters in DecideNet
	15	(https://arxiv.org/pdf/1712.06679.pdf)
	16	It does not implement decide net correctly
	17	Just 'nearly' correct to estimate number of parameters
	18	"""
	19	def __init__(self, load_weights=False):
	20	super(DecideNet, self).__init__()
	21	self.seen = 0
	22	self.reg_net_feat = [(7, 20, 1), (5, 40, 1), (5, 20, 1), (5, 10, 1), (1, 1, 1)]
	23	self.reg_net = make_layers(self.reg_net_feat, 3)
	24	self.quality_net_feat = [(7, 20, 1), (5, 40, 1), (5, 20, 1), (1, 1, 1)]
	25	self.quality_net = make_layers(self.quality_net_feat, 3)
	26	# backbone = torchvision.models.mobilenet_v2(pretrained=False).features
	27	backbone = torchvision.models.detection.backbone_utils.resnet_fpn_backbone('resnet101','False')
	28	# model = torchvision.models.detection.fasterrcnn_resnet50_fpn()
	29	# backbone.out_channels = 1280
	30	anchor_generator = AnchorGenerator(sizes=((32, 64, 128, 256, 512),),
	31	aspect_ratios=((0.5, 1.0, 2.0),))
	32	roi_pooler = torchvision.ops.MultiScaleRoIAlign(featmap_names=[0],
	33	output_size=7,
	34	sampling_ratio=2)
	35	self.fastercnn = FasterRCNN(backbone,
	36	num_classes=2,
	37	rpn_anchor_generator=anchor_generator,
	38	box_roi_pool=roi_pooler)
	39
	40
	41
	42	def forward(self, x):
	43	x = self.frontend(x)
	44	x = self.backend(x)
	45	x = self.output_layer(x)
	46
	47	# remove channel dimension
	48	# (N, C_{out}, H_{out}, W_{out}) => (N, H_{out}, W_{out})
	49	x = torch.squeeze(x, dim=1)
	50	return x
	51
	52	# def _initialize_weights(self):
	53	# for m in self.modules():
	54	# if isinstance(m, nn.Conv2d):
	55	# nn.init.normal_(m.weight, std=0.01)
	56	# if m.bias is not None:
	57	# nn.init.constant_(m.bias, 0)
	58	# elif isinstance(m, nn.BatchNorm2d):
	59	# nn.init.constant_(m.weight, 1)
	60	# nn.init.constant_(m.bias, 0)
	61
	62
	63	def make_layers(cfg, in_channels=3, batch_norm=False):
	64	layers = []
	65	for v in cfg:
	66	if v is 'M':
	67	layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
	68	else:
	69	kernel_size = v[0]
	70	n_filter = v[1]
	71	dilated_rate = v[2]
	72	conv2d = nn.Conv2d(in_channels, n_filter, kernel_size=kernel_size, padding=dilated_rate, dilation=dilated_rate)
	73	if batch_norm:
	74	layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU(inplace=False)]
	75	else:
	76	layers += [conv2d, nn.ReLU(inplace=False)]
	77	in_channels = n_filter
	78	return nn.Sequential(*layers)
	79
	80	if __name__ == "__main__":
	81	net = DecideNet()
	82	n_param = very_simple_param_count(net)
	83	print(n_param)
	84	resnet101 = torchvision.models.detection.backbone_utils.resnet_fpn_backbone('resnet101','False')
	85	print(very_simple_param_count(resnet101))

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