poppinace / libranet Goto Github PK
View Code? Open in Web Editor NEW(ECCV'20) Weighing Counts: Sequential Crowd Counting by Reinforcement Learning
License: Other
(ECCV'20) Weighing Counts: Sequential Crowd Counting by Reinforcement Learning
License: Other
Hi,
Thanks for your great work. For the cross-dataset evaluation, did you use the same model trained on the source domain for cross-dataset evaluation?
Is there an ETA on when the code will be coming @poppinace? Using RL for this is awesome, and I'm keen to have a look under the hood. For what it's worth, an open license would be great too.
Hi, can you explain more on the ground truth file (.csv in your preprocessed training dataset)? From my understanding, you divide the whole image into 32*32 patches and corresponding count value. So each element in the csv file represents its patch count and its patch relative location. Am I correct?
For example:
I use $ cat 9.csv
and it shows:
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
1,5,14,20,23,16,5,1,10,1,12,23,32,36,40,42
37,37,40,42,39,39,44,44,40,42,39,45,45,47,42,44
26,29,20,20,19,23,23,27,22,22,29,27,32,32,30,30
if recycle_ind < parameters['ACTION_NUMBER'] - 1:
start_mask_random = ( (count_rem + net.A[recycle_ind] >= 0) & (start_ind_random == -1) )
start_ind_random[start_mask_random] = recycle_ind
**end_mask_random** = ( count_rem + net.A[recycle_ind] < parameters['Interval_N'] )
**end_ind_random**[end_mask_random] = recycle_ind
maskselect_end = (sort[recycle_ind]==parameters['ACTION_NUMBER']-1)
action_sort = sort[recycle_ind]
A_sort = np.squeeze(net.A_mat[action_sort])
_ind_max = (( (count_rem + A_sort < parameters['Interval_N']) & (count_rem + A_sort >= 0) | maskselect_end) & (mask_max_find==0) ) & (mask_last==0)
action_max[_ind_max] = action_max[_ind_max] + sort[recycle_ind] [_ind_max] #?
mask_max_find = mask_max_find + ( (count_rem + A_sort < parameters['Interval_N']) & (count_rem + A_sort >= 0) | maskselect_end ).astype(np.int8)
action_random = (start_ind_random + (end_ind_random + 2 - start_ind_random ) * np.random.rand(h, w)).astype(np.int8)
Why don't we generate action_random at random, but start_ ind_ Random and end_ ind_ random?
Thanks for releasing the code for inspiring work!
Do you have preprocessed version for the SHT Part_B dataset?
Thanks in advance!
Joseph
Hi there, could you please add a standard license? Thanks!
I notice that your train dataset is not the standard Shanghaitech part-A dataset. You do some data augmentation. Could you share some details about augmentation and what the CSV files mean?
Thanks!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.