The main challenge would be the spatial and temporal coverage of the data. If you have a dedicated research team or a set of traffic sensors that can be employed, then this approach could be valid. Otherwise, you would have to restrict yourself to the typical peak hours. This is still valid for many types of analysis, but it would not be possible to generalize and extend it beyond the scope of work.

If you could include the region of your study, some of the other researchers may have that data which they may like to help you with.

Uneb Gazder

Thank you for your answer and suggestions.

For gate counting, I've been granted access to 8 city CCTV cameras on public streets positioned at major junctions around my study area in the central business district of Bangkok. I've collected footage from 3 weekdays and 3 weekend days × 8 cameras × time intervals, covering various periods to accurately capture traffic patterns.

I've already implemented a YOLO model with custom data annotations specifically for Bangkok's street context—including dense motorcycle traffic, mini van taxis, tuk-tuks, other local vehicles, and, of course, bicycles.

Now, I want to perform gate counting by tracking when the center of vehicle bounding boxes crosses lines I've drawn on the video feeds from all 8 cameras. This approach allows me to collect traffic counts from 8 key locations in the district, providing a comprehensive picture of traffic volume throughout the area.

This is why I want to pursue this direction to obtain trip distribution or synthetic origin-destination data for bicycles.

Kulacha Sirikhan

That's great. In this case, I believe you have a valid approach and sufficient dataset. I wish you all the best for your research. If you ever need to collaborate on anything related to transportation then feel free to contact me.