ESVR 0 Posted August 11, 2022 I'm working on a facial recognition project where I'm trying to estimate Age and Gender of a person from a video feed. I trained a model and inferences on my development machine is taking around 200ms. I converted these models to tflite and pushed them to Tinker Edge T and the inferences are pretty slow. The average inference time is around 5-6 sec. I'd like to get it down to 200ms just like in my development machine and would appreciate any help/guidance from the community. Does GPU/TPU are enabled automatically when i run the program or I have activate them manually? Not sure if this is the problem or something else. Please help me Share this post Link to post Share on other sites
yi-hsin_hung 3 Posted August 16, 2022 Hi @ESVR: Did you have convert the model by edgetpu_compiler for edgetpu tflite model. Thank a lot! https://coral.ai/docs/edgetpu/compiler/ Share this post Link to post Share on other sites
