Part Number: TDA4VM
Hi ：
We have some question when use TDA4，as follows:
1. Centernet human pose detection network is converted to onnx model after using QAT training (the backbone network used is resnet34 provided by Centernet official website), there is no output result after using TIDL8.0 inference, and I set quantizationStyle = 3 according to the official guidance configuration file, quantization. calibrationOption=64.
2. Centernet human pose detection network uses the PTQ quantification model, and the prediction points have deviations, and the model is relatively large。
Could you tell me how to solve it？