關於kl520輸出的post processing
不好意思,我有些疑問,由於許多模型需要做一些後處理來得到bounding boxes, score, classes等等資訊,但在https://github.com/kneron/host_lib/blob/master/python/examples_kl520/cam_dme_serial_ssd_fd.py的例子中,似乎不需要輸出整個feature maps,而是直接得到post processing後的結果。
所以我想問的是
1.要如何達到這樣的結果,是轉換成nef的過程中有需要額外的設定嗎?
2.只有特定model才能自動完成post processing嗎?那yolov5呢?
The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.
Comments
您好,
您提到的project (host_lib)裡面也有postprocessing。dme_inference裡面會呼叫dme_get_result,在那裡面會拿raw output (feature map)去做postprocessing:
Host_lib裡的postprocessing範例在host_lib/python/common/pre_post_process裡面。