Thanks! I have implemented it as follows. It is working, but the detection is not correct. The model I'm using is working and has been validated with still image. Is there a simple way to check if the resulting image is correct (e.g. save as jpg)?
Code:
void TfStage::runInference(){int input = interpreter_->inputs()[0];StreamInfo tf_info;tf_info.width = tf_w_, tf_info.height = tf_h_, tf_info.stride = tf_w_ * 3;std::vector<uint8_t> inference_image;if (rgb_input){inference_image = Yuv420ToRgb(lores_copy_.data(), lores_info_, tf_info);}else{inference_image.resize(lores_info_.width * lores_info_.height);std::copy(lores_copy_.data(), lores_copy_.data() + inference_image.size(), inference_image.begin());}...}Statistics: Posted by gftrobots — Mon Jul 22, 2024 2:54 pm