Notifications
Clear all

Question Enquiries on conditional counting for dabble.statistics with PeekingDuck

3 Posts
2 Users
0 Likes
273 Views
 Vkoh
(@vkoh)
New Member Member
Joined: 9 months ago
Posts: 2
Topic starter  

Hi! I would like to ask if there are any ways to count the specific number of each type of object detected by the yolo model?

I tried to use dabble.tracking followed by conditional counting in dabble.statistics. However, I realised that the 'obj_attrs'  returned only has the 'ids' attribute unlike the example (in the table under dabble.statistics) shown in the documentation where the 'details' attribute is also present. I am unable to find any other nodes in dabble with a suitable attribute . Thus I would like to ask if there are any ways to work around this or am I missing out on something here. 

Thank you for your time!

This topic was modified 9 months ago by Vkoh

   
Quote
Gao Hongnan
(@gao-hongnan)
Eminent Member Moderator
Joined: 1 year ago
Posts: 28
 

Hi, after discussing with my team, it cannot be done within the current PeekingDuck codebase, unfortunately.

You may consider writing a custom node to process the outputs of `model.yolo`. The `ids` in `dabble.statistics` tutorial are generated by `dabble.tracking`, which are not the same as the objects' class ids. PeekingDuck models do not output `obj_attrs` (used by `dabble.statistics`); The custom node can include self defined logic to process `bbox_labels` and output a group by summary strong, which canbe displayed by `draw.legend`.


   
ReplyQuote
 Vkoh
(@vkoh)
New Member Member
Joined: 9 months ago
Posts: 2
Topic starter  

@gao-hongnan Thank you for the reply!


   
ReplyQuote
Share: