peterjc123
peterjc123
@gj-raza The schema of the `TFLITE_DETECTION_POSTPROCESS` op is added [here](https://github.com/alibaba/TinyNeuralNetwork/commit/bd70e36e6cfa301c605ae3bc42db4eda159a183d#diff-4d917771c0f7ad00a838050a2ecc673f4ebadd28ec480330a5e4a1b905eb9c36R12) in case you may need that.
@gj-raza Yes, you'll need to do steps 1-6 as I described.
Not going to implement it at our side because it is rarely used in our scenarios. However, if you have any questions implementing this feature, you are free to ask...
据我所知，TFLite模型不支持动态输入大小，只能在AllocateTensors前用ResizeInput来一次性的改变输入的大小。如你有支持动态维度输入的TFLite模型，麻烦提供一个给我们，我们可以研究下怎么实现。
> PyTorch 动态 shape 模型的话似乎必须要用 `torch.jit.script` 来导出？ PyTorch这边其实无所谓，只要TFLite能支持，我们这边可以添加接口，通过一个pass来传播shape的动态性就可以了
> > 据我所知，TFLite模型不支持动态输入大小，只能在AllocateTensors前用ResizeInput来一次性的改变输入的大小。如你有支持动态维度输入的TFLite模型，麻烦提供一个给我们，我们可以研究下怎么实现。 > > 您好，关于TFLite模型是否支持动态输入大小这个问题，我了解不深，只是参阅了 https://stackoverflow.com/questions/55701663/input-images-with-dynamic-dimensions-in-tensorflow-lite/55732431#55732431 和 [tensorflow/tensorflow#41807](https://github.com/tensorflow/tensorflow/issues/41807) 这两个链接的回答 你试下你的模型走resize_input_tensor可以正常工作吗？
嗯，那reshape算子都得特殊处理下，你看下导出的torchscript包不包含shape相关的操作？如果有相关操作的话，我们只需要把shape跟踪下来就行了，如果没有就比较复杂了
> 1. Unify the quantization parameters of (Y, B) to the quantization parameters of A and also we need to disable the observers in those variables. This is simpler I...
@ogoffart Any update on this? As a desktop application, popups are often used to show info or fatal error messages. Without that, we can only use status bar, which leads...
@ogoffart It seems that a Dialog still needs to wrapped in a PopupWindow, so that those issues still exist.