We are using the person-detection-retail-0013.xml model. What we are seeing is that if the full body is not visible (i.e. the legs are not visible as hidden behind counter) then the object is not detected.
Is there a replacement model that is more recommended or a way to change parameters to still identify if the legs are hidden/portions of the body are hidden?
Thank you, appreciate it.
Is there though some documentation to help decide or is it try and see sort of effort?
The views are from security cameras, so at an upwards angle (like 45 degrees facing down to the person) and at times the person's legs are hidden due to physical items on site.
@Duggy for Intel pre-trained models usually internal datasets used at training time, but for some models we have published training pipeline , so you may check the link to see what are training dataset used for these models and evaluate if it does fit your requirements, or fine-tune model for your specific needs (see person-detection-02xx models description).
This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.