Closed kitterive closed 1 year ago
I convert keypoints from tf pb model to onnx, and found the result is not same pb model result.
可能是由于keypoints和 detect 两个模型都有traning flag的原因,在导出模型时,能去掉training flag这个输入吗,设为固定值False, 这样就一个输入,比较适合推理
solved
请问下您是怎么解决的呢?tf2onnx的时候能设置固定输入吗?有没有转换脚本呢
I convert keypoints from tf pb model to onnx, and found the result is not same pb model result.