yuhangsong / dhp Goto Github PK

I've run the YUV preprocessing code by modified mode = 'data_processor' and data_processor_id = 'mp4_to_yuv' in config.py.
There were lots of YUV files generated in the location of PVS dataset, but there's no '1.yuv' file in 'results/reproduce_17/temp/get_view/w_//1.yuv.

Did I missed any step in data preprocessing?

哈囉，
我在跑 training offline 的時候會遇到 No such file or directory: '../results/reproduce_17/temp/get_view/w_5/Bicycle/1.yuv' 的問題，
get_view/w_*/影片名稱/ 底下也確實沒有 1.yuv 檔案，
想請問 1.yuv 是 preprocessing 的時候應該被 generate 的嗎？
preprocessing 時在 dataset 內generate 出來的 '影片名稱.yuv' 檔案跟 1.yuv 是同一個嗎？

数据集

您好，我点开那个DropBox显示 “/Yuhang Song/Dataset”文件夹不存在。点开那个EM数据库好像也只有一个README文件，请问是没有传上去吗？

Update to md.

@MinglangQiao Please update your name & photo in the md. Besides, update the dataset detail description part, and update the pointer to your EM git repo.

On 'v' action rewards

Hello dear mr. Yuhang Song,

In the paper, it is mentioned that the rewards for action v are given by

And the parameters θv are optimized according to the rule:

In the code,

DHP/envs.py

Lines 488 to 502 in 73ddec2

 '''get direction reward and ground-truth v from data_base in last state''' 

 last_prob, distance_per_data = suppor_lib.get_prob( 

 lon=self.last_lon, 

 lat=self.last_lat, 

 theta=action * 45.0, 

 subjects=self.subjects, 

 subjects_total=self.subjects_total, 

 cur_data=self.last_data, 

 ) 

 '''rescale''' 

 distance_per_step = distance_per_data * self.data_per_step 

 '''convert v to degree''' 

 degree_per_step = distance_per_step / math.pi * 180.0 

 '''set v_lable''' 

 v_lable = degree_per_step

there seems to be no reward for v calculated, instead v_lable is estimated as a "weighted" target value (sum of subject_i_v * similarity),

DHP/suppor_lib.py

Lines 154 to 159 in 73ddec2

 distance_per_data = 0.0 

 for i in range(subjects_total): 

 if config.if_normalize_v_lable: 

 distance_per_data += prob_dic_normalized[i] * subjects[i].data_frame[cur_data].v 

 else: 

 distance_per_data += prob_dic[i] * subjects[i].data_frame[cur_data].v

which then contributes another term (v-v_lable)^2 in the loss function:

DHP/a3c.py

Lines 238 to 239 in 73ddec2

 # v loss 

 v_loss = 0.5 * tf.reduce_sum(tf.square(pi.v - self.v_lable))

Is there any particular reason why the direct sum of rewards is not calculated, and instead the above approach is considered?

online main part

see here

run online-DHP

Hi, I am interested in your online-DHP newwork. Introduced by your Readme, I have known how to use the offline DHP. Therefore, can you write the method for online DHP method?

yuhangsong / dhp Goto Github PK

dhp's Issues

Can you provide the HTC VIVE SDK for tracking head movements or its download address?Thank you very much！

online_DHP

Save off line map

Log CC and heatmap.

How to change the code to GPU.

Data Preprocessing file missing

数据集

Update to md.

On 'v' action rewards

online main part

run online-DHP

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	'''get direction reward and ground-truth v from data_base in last state'''
	last_prob, distance_per_data = suppor_lib.get_prob(
	lon=self.last_lon,
	lat=self.last_lat,
	theta=action * 45.0,
	subjects=self.subjects,
	subjects_total=self.subjects_total,
	cur_data=self.last_data,
	)
	'''rescale'''
	distance_per_step = distance_per_data * self.data_per_step
	'''convert v to degree'''
	degree_per_step = distance_per_step / math.pi * 180.0
	'''set v_lable'''
	v_lable = degree_per_step

	distance_per_data = 0.0
	for i in range(subjects_total):
	if config.if_normalize_v_lable:
	distance_per_data += prob_dic_normalized[i] * subjects[i].data_frame[cur_data].v
	else:
	distance_per_data += prob_dic[i] * subjects[i].data_frame[cur_data].v

	# v loss
	v_loss = 0.5 * tf.reduce_sum(tf.square(pi.v - self.v_lable))