Closed jayle19930918 closed 4 years ago
Thank you for your work! I compiled the model using TVM at Windows10 successfully and got the correct bouding box of face. But it needs 1200ms to detect one frame which is so strange. I do no know what is the problem.
It could be the problems:
you can check out the listed issues.
Thank you for your work! I compiled the model using TVM at Windows10 successfully and got the correct bouding box of face. But it needs 1200ms to detect one frame which is so strange. I do no know what is the problem.
It could be the problems:
- tvm complier target.
- Hardware limitation. I'm using MBP 15' as testing env.
- Too large input size
you can check out the listed issues.
Thanks for your suggestions! The pretrained model I compiled was RetinaFace-R50, which is a medium size model with ResNet50 backbone. Then I found the MobileNet0.25 model and compiled it. Now it works well!
Thank you for your work! I compiled the model using TVM at Windows10 successfully and got the correct bouding box of face. But it needs 1200ms to detect one frame which is so strange. I do no know what is the problem.