@alexisv098 8 Hi, usually with Tiny Yolo the post processing consists of filtering out various objects that do not meet your criteria. For example, objects with low scores and any duplicate boxes. With the model that I mentioned at https://ncsforum.movidius.com/discussion/comment/2555/#Comment_2555, you will end up with an array of 21125 items. Why 21125? This model splits up the input image into 13 x 13 grid with 5 anchor boxes per grid cell and then for each of those 5 bounding boxes, there are 20 classes predictions with 4 bounding box coordinates and 1 object score (13 x 13 x 5 x 25 = 21155). So the array will have 21125 results.
You can see an example of this in this project lines 136 through lines 168.
I see that your model returns an array of 18750 elements. Where did you get this model from? Can you provide a link to your model and your code? Thanks.
It is likely just a matter of finding out how your model is configured. As you can see here, the ncappzoo's Tiny Yolo V1 model splits up the image into 7x7 grid with 2 anchor boxes per grid and 20 classes. The post processing is a little bit different, but a lot of it is the same idea (filter scores lower than a specific threshold, sort).
