mx/Jungfraujoch
2
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 67 Wiki Activity

* Enhancements for XFEL

Browse Source 
* Enhancements for EIGER
* Writer is more flexible and capable of handling DECTRIS data
This commit is contained in:
leonarski_f
2024-03-05 20:41:47 +01:00
parent 71d862b706
commit d315506633
165 changed files with 5440 additions and 2230 deletions
Download Patch File Download Diff File Expand all files Collapse all files
resonet/NeuralNetResPredictor.cpp
+20 -19
View File
@@ -5,27 +5,21 @@
#include <cmath>
#include "../common/JFJochException.h"
NeuralNetResPredictor::NeuralNetResPredictor(const DiffractionExperiment &in_experiment)
: experiment(in_experiment), model_input(512*512)
NeuralNetResPredictor::NeuralNetResPredictor(const std::string& model_path)
: model_input(512*512)
#ifdef JFJOCH_USE_TORCH
, device(torch::kCUDA)
#endif
{
float max_direction = std::max(in_experiment.GetXPixelsNum(), in_experiment.GetYPixelsNum()) / 2.0;
pool_factor = std::lround(max_direction / 512.0f);
if (pool_factor <= 0)
throw JFJochException(JFJochExceptionCategory::InputParameterInvalid, "Detector size is too small");
if (pool_factor > 8)
throw JFJochException(JFJochExceptionCategory::InputParameterInvalid, "Detector size is too large");
#ifdef JFJOCH_USE_TORCH
module = torch::jit::load(experiment.GetNeuralNetModelPath());
module = torch::jit::load(model_path);
module.to(device);
#endif
}
template<class T>
void NeuralNetResPredictor::PrepareInternal(const T* image) {
void NeuralNetResPredictor::PrepareInternal(const DiffractionExperiment& experiment, const T* image) {
size_t pool_factor = GetMaxPoolFactor(experiment);
size_t xpixel = experiment.GetXPixelsNum();
size_t ypixel = experiment.GetYPixelsNum();
size_t start_x = std::lround(experiment.GetBeamX_pxl());
@@ -52,23 +46,30 @@ void NeuralNetResPredictor::PrepareInternal(const T* image) {
}
}
void NeuralNetResPredictor::Prepare(const int16_t *image) {
PrepareInternal(image);
void NeuralNetResPredictor::Prepare(const DiffractionExperiment& experiment, const int16_t *image) {
PrepareInternal(experiment, image);
}
void NeuralNetResPredictor::Prepare(const int32_t *image) {
PrepareInternal(image);
void NeuralNetResPredictor::Prepare(const DiffractionExperiment& experiment, const int32_t *image) {
PrepareInternal(experiment, image);
}
size_t NeuralNetResPredictor::GetMaxPoolFactor() const {
size_t NeuralNetResPredictor::GetMaxPoolFactor(const DiffractionExperiment& experiment) const {
float max_direction = std::max(experiment.GetXPixelsNum(), experiment.GetYPixelsNum()) / 2.0;
size_t pool_factor = std::lround(max_direction / 512.0f);
if (pool_factor <= 0)
throw JFJochException(JFJochExceptionCategory::InputParameterInvalid, "Detector size is too small");
if (pool_factor > 8)
throw JFJochException(JFJochExceptionCategory::InputParameterInvalid, "Detector size is too large");
return pool_factor;
}
float NeuralNetResPredictor::Inference(const void *image) {
float NeuralNetResPredictor::Inference(const DiffractionExperiment& experiment, const void *image) {
if (experiment.GetPixelDepth() == 2)
Prepare((int16_t *) image);
Prepare(experiment, (int16_t *) image);
else
Prepare((int32_t *) image);
Prepare(experiment, (int32_t *) image);
#ifdef JFJOCH_USE_TORCH
auto options = torch::TensorOptions().dtype(at::kFloat);