Open ijarraki opened 4 years ago
Hi!
You can try this project, it runs in real time on my laptop:
./MultitargetTracker
Advanced settings you can set here: https://github.com/Smorodov/Multitarget-tracker/blob/master/example/examples.h#L618
In this project now don't used re-id and another deep-embeddings but I can to add they.
Hopefully "re-id" won't be an issue, this is a matter of simply tracking "object N" across the field-of-view (no occlusion) and count the number of seconds the object/centroid stays within each of three defined regions-of-interest and then log this to sql/sqlite/csv. The reason I brought this up for this project is the "abandoned detector" mentioned in the README.
I'm not sure how I would achieve this using this code, any ideas?
Abandoned detector used for luggage, I'm implemented it 2-3 years ago for one issue: https://www.youtube.com/watch?v=fpkHRsFzspA But it not tested. You can try it: settings.m_useAbandonedDetection = true;
This is not a bug but rather a question/feature request. I've been looking for some code to achieve this (what i intially thought would be) simple task:
1: Monitor video input stream 2: Classify objects entering view (e.g. car/not car). 3: Track the "car" as it moves across the input 4: Time how long the "car" stayed in each of say three predefined regions of interest, log sqlite/sql/csv.
Some time ago I came across this project: https://github.com/Ashwin-Ramesh2607/DeepSort-YOLOv3 but I've found it to be painfully slow, among other things since it's hardcoded for high resolution input. Is this something that could be achieved using this project instead, and if so: how?