Calculating MWIS becomes slow

[ijunglee]

Hi, I have tried this code on my dataset which has 100 frames. The maximum observations is 4. I noticed that the calculating MWIS procedure becames really slow after frame 49. Here are the parameters I set:

image_area = 422288 # Image width x height in pixels
gating_area = 1 # Gating area for new detections
k = 0 # Gain or blending factor
q = 0.00001 # Kalman filter process variance
r = 0.01 # Estimate of measurement variance
n = 1 # N-scan pruning parameter

Would you suggest me change any parameters to make it run faster? Thank you so much!

[jonperdomo]

Hi, I will be implementing other parameters soon that should speed it up, see issue #7

[ijunglee]

Hi, I will be implementing other parameters soon that should speed it up, see issue #7

Thank you so much. Look forward to the update.

I also want to know how to set the parameters of gating area, k, q, r, and n.
I tried this test01.csv with one or two subjects in a frame, and the total is 60 frames.
test01.csv
The parameters I used are as follows:
image_area = 422288 # Image width x height in pixels
gating_area = 1 # Gating area for new detections
k = 1 # Gain or blending factor
q = 0.00001 # Kalman filter process variance
r = 0.1 # Estimate of measurement variance
n = 1 # N-scan pruning parameter

However, I got track 0 to 7 in my output file, which is weird.
test01_output.csv

Do you have any ideas how to solve this problem? Thanks a lot!

[jonperdomo]

Hi, below are more details on the parameters, and please browse the paper for specifics on tuning these. I would try increasing K if your measurements have high accuracy, and increasing N for a higher accuracy solution (but slower runtime) as well. I am currently working on adding the Bth and Nmiss parameters which should improve performance as well. I'm also adding the parameters as command-line arguments so you can test different values more easily.

Parameters

Modify parameters by editing the file params.txt:

Parameter	Description
v	The image (frame) area in pixels (Default: 307200). The likelihood under the null hypothesis for an observation becomes the probability of detection (PD) 1/V.
dth	Gating area for new detections implemented as the threshold for the Mahalinobis distance d2 between the observation and prediction (Default=1000).

Kalman filter parameters:

Parameter	Description
k	Gain or blending factor. Higher gain results in a greater influence of the measurement relative to the filter's prediction (Default=0)
q	Initial estimate of the process noise covariance (Default=0.00001)
r	Initial estimate of the measurement noise covariance (Default=0.01)

Track tree pruning parameters:

Parameter	Description
n	Go back N frames and prune branches that diverge from the solution. Larger N yields a more accurate solution due to a larger window, but will take a longer time (Default=1).
bth	If the number of branches exceeds the number Bth, then prune the track tree to only retain the top Bth branches.
nmiss	A track hypothesis is deleted if it reaches Nmiss consecutive frames of missing observations, which are due to occlusion or a false negative.

[jonperdomo]

I will test with your inputs as well

[Engineero]

Any updates on speeding up the MWIS calculation? Are the new parameters included, and if so how do I specify them?

[jonperdomo]

Still working on this, sorry!

[jonperdomo]

@Engineero I have added two missing pruning parameters from the paper: Bth for setting the maximum number of branches, and Nmiss for the maximum number of consecutive frames with missing observations for a track in commit 5627c53. Please try adjusting these parameters to improve performance.