public static (Image[] Halves1, Image[] Halves2, float2[] Stats) TrainOnVolumes(NoiseNet3D network,
Image[] halves1,
Image[] halves2,
Image[] masks,
float angpix,
float lowpass,
float upsample,
bool dontFlatten,
bool performTraining,
int niterations,
float startFrom,
int batchsize,
int gpuprocess,
Action <string> progressCallback)
{
if (batchsize != 4)
{
niterations = niterations * 4 / batchsize;
Debug.WriteLine($"Adjusting the number of iterations to {niterations} to match different batch size.\n");
}
GPU.SetDevice(gpuprocess);
#region Mask
Debug.Write("Preparing mask... ");
progressCallback?.Invoke("Preparing mask... ");
int3[] BoundingBox = Helper.ArrayOfFunction(i => new int3(-1), halves1.Length);
if (masks != null)
{
for (int i = 0; i < masks.Length; i++)
{
Image Mask = masks[i];
Mask.TransformValues((x, y, z, v) =>
{
if (v > 0.5f)
{
BoundingBox[i].X = Math.Max(BoundingBox[i].X, Math.Abs(x - Mask.Dims.X / 2) * 2);
BoundingBox[i].Y = Math.Max(BoundingBox[i].Y, Math.Abs(y - Mask.Dims.Y / 2) * 2);
BoundingBox[i].Z = Math.Max(BoundingBox[i].Z, Math.Abs(z - Mask.Dims.Z / 2) * 2);
}
return(v);
});
if (BoundingBox[i].X < 2)
{
throw new Exception("Mask does not seem to contain any non-zero values.");
}
BoundingBox[i] += 64;
BoundingBox[i].X = Math.Min(BoundingBox[i].X, Mask.Dims.X);
BoundingBox[i].Y = Math.Min(BoundingBox[i].Y, Mask.Dims.Y);
BoundingBox[i].Z = Math.Min(BoundingBox[i].Z, Mask.Dims.Z);
}
}
Console.WriteLine("done.\n");
#endregion
#region Load and prepare data
Console.WriteLine("Preparing data:");
List <Image> Maps1 = new List <Image>();
List <Image> Maps2 = new List <Image>();
List <Image> HalvesForDenoising1 = new List <Image>();
List <Image> HalvesForDenoising2 = new List <Image>();
List <float2> StatsForDenoising = new List <float2>();
for (int imap = 0; imap < halves1.Length; imap++)
{
Debug.Write($"Preparing map {imap}... ");
progressCallback?.Invoke($"Preparing map {imap}... ");
Image Map1 = halves1[imap];
Image Map2 = halves2[imap];
float MapPixelSize = Map1.PixelSize / upsample;
if (!dontFlatten)
{
Image Average = Map1.GetCopy();
Average.Add(Map2);
if (masks != null)
{
Average.Multiply(masks[imap]);
}
float[] Spectrum = Average.AsAmplitudes1D(true, 1, (Average.Dims.X + Average.Dims.Y + Average.Dims.Z) / 6);
Average.Dispose();
int i10A = Math.Min((int)(angpix * 2 / 10 * Spectrum.Length), Spectrum.Length - 1);
float Amp10A = Spectrum[i10A];
for (int i = 0; i < Spectrum.Length; i++)
{
Spectrum[i] = i < i10A ? 1 : (Amp10A / Math.Max(1e-10f, Spectrum[i]));
}
Image Map1Flat = Map1.AsSpectrumMultiplied(true, Spectrum);
Map1.FreeDevice();
Map1 = Map1Flat;
Map1.FreeDevice();
Image Map2Flat = Map2.AsSpectrumMultiplied(true, Spectrum);
Map2.FreeDevice();
Map2 = Map2Flat;
Map2.FreeDevice();
}
if (lowpass > 0)
{
Map1.Bandpass(0, angpix * 2 / lowpass, true, 0.01f);
Map2.Bandpass(0, angpix * 2 / lowpass, true, 0.01f);
}
if (upsample != 1f)
{
Image Map1Scaled = Map1.AsScaled(Map1.Dims * upsample / 2 * 2);
Map1.FreeDevice();
Map1 = Map1Scaled;
Map1.FreeDevice();
Image Map2Scaled = Map2.AsScaled(Map2.Dims * upsample / 2 * 2);
Map2.FreeDevice();
Map2 = Map2Scaled;
Map2.FreeDevice();
}
Image ForDenoising1 = Map1.GetCopy();
Image ForDenoising2 = Map2.GetCopy();
if (BoundingBox[imap].X > 0)
{
Image Map1Cropped = Map1.AsPadded(BoundingBox[imap]);
Map1.FreeDevice();
Map1 = Map1Cropped;
Map1.FreeDevice();
Image Map2Cropped = Map2.AsPadded(BoundingBox[imap]);
Map2.FreeDevice();
Map2 = Map2Cropped;
Map2.FreeDevice();
}
float2 MeanStd = MathHelper.MeanAndStd(Helper.Combine(Map1.GetHostContinuousCopy(), Map2.GetHostContinuousCopy()));
Map1.TransformValues(v => (v - MeanStd.X) / MeanStd.Y);
Map2.TransformValues(v => (v - MeanStd.X) / MeanStd.Y);
ForDenoising1.TransformValues(v => (v - MeanStd.X) / MeanStd.Y);
ForDenoising2.TransformValues(v => (v - MeanStd.X) / MeanStd.Y);
HalvesForDenoising1.Add(ForDenoising1);
HalvesForDenoising2.Add(ForDenoising2);
StatsForDenoising.Add(MeanStd);
GPU.PrefilterForCubic(Map1.GetDevice(Intent.ReadWrite), Map1.Dims);
Map1.FreeDevice();
Maps1.Add(Map1);
GPU.PrefilterForCubic(Map2.GetDevice(Intent.ReadWrite), Map2.Dims);
Map2.FreeDevice();
Maps2.Add(Map2);
Debug.WriteLine(" Done.");
}
if (masks != null)
{
foreach (var mask in masks)
{
mask.FreeDevice();
}
}
#endregion
int Dim = network.BoxDimensions.X;
progressCallback?.Invoke($"0/{niterations}");
if (performTraining)
{
GPU.SetDevice(gpuprocess);
#region Training
Random Rand = new Random(123);
int NMaps = Maps1.Count;
int NMapsPerBatch = Math.Min(128, NMaps);
int MapSamples = batchsize;
Image[] ExtractedSource = Helper.ArrayOfFunction(i => new Image(new int3(Dim, Dim, Dim * MapSamples)), NMapsPerBatch);
Image[] ExtractedTarget = Helper.ArrayOfFunction(i => new Image(new int3(Dim, Dim, Dim * MapSamples)), NMapsPerBatch);
for (int iter = (int)(startFrom * niterations); iter < niterations; iter++)
{
int[] ShuffledMapIDs = Helper.RandomSubset(Helper.ArrayOfSequence(0, NMaps, 1), NMapsPerBatch, Rand.Next(9999));
for (int m = 0; m < NMapsPerBatch; m++)
{
int MapID = ShuffledMapIDs[m];
Image Map1 = Maps1[MapID];
Image Map2 = Maps2[MapID];
int3 DimsMap = Map1.Dims;
int3 Margin = new int3((int)(Dim / 2 * 1.5f));
//Margin.Z = 0;
float3[] Position = Helper.ArrayOfFunction(i => new float3((float)Rand.NextDouble() * (DimsMap.X - Margin.X * 2) + Margin.X,
(float)Rand.NextDouble() * (DimsMap.Y - Margin.Y * 2) + Margin.Y,
(float)Rand.NextDouble() * (DimsMap.Z - Margin.Z * 2) + Margin.Z), MapSamples);
float3[] Angle = Helper.ArrayOfFunction(i => new float3((float)Rand.NextDouble() * 360,
(float)Rand.NextDouble() * 360,
(float)Rand.NextDouble() * 360) * Helper.ToRad, MapSamples);
{
ulong[] Texture = new ulong[1], TextureArray = new ulong[1];
GPU.CreateTexture3D(Map1.GetDevice(Intent.Read), Map1.Dims, Texture, TextureArray, true);
//Map1.FreeDevice();
GPU.Rotate3DExtractAt(Texture[0],
Map1.Dims,
ExtractedSource[m].GetDevice(Intent.Write),
new int3(Dim),
Helper.ToInterleaved(Angle),
Helper.ToInterleaved(Position),
(uint)MapSamples);
//ExtractedSource[MapID].WriteMRC("d_extractedsource.mrc", true);
GPU.DestroyTexture(Texture[0], TextureArray[0]);
}
{
ulong[] Texture = new ulong[1], TextureArray = new ulong[1];
GPU.CreateTexture3D(Map2.GetDevice(Intent.Read), Map2.Dims, Texture, TextureArray, true);
//Map2.FreeDevice();
GPU.Rotate3DExtractAt(Texture[0],
Map2.Dims,
ExtractedTarget[m].GetDevice(Intent.Write),
new int3(Dim),
Helper.ToInterleaved(Angle),
Helper.ToInterleaved(Position),
(uint)MapSamples);
//ExtractedTarget.WriteMRC("d_extractedtarget.mrc", true);
GPU.DestroyTexture(Texture[0], TextureArray[0]);
}
//Map1.FreeDevice();
//Map2.FreeDevice();
}
float[] PredictedData = null, Loss = null;
{
float CurrentLearningRate = 0.0001f * (float)Math.Pow(10, -iter / (float)niterations * 2);
for (int m = 0; m < ShuffledMapIDs.Length; m++)
{
int MapID = m;
bool Twist = Rand.Next(2) == 0;
if (Twist)
{
network.Train(ExtractedSource[MapID].GetDevice(Intent.Read),
ExtractedTarget[MapID].GetDevice(Intent.Read),
CurrentLearningRate,
0,
out PredictedData,
out Loss);
}
else
{
network.Train(ExtractedTarget[MapID].GetDevice(Intent.Read),
ExtractedSource[MapID].GetDevice(Intent.Read),
CurrentLearningRate,
0,
out PredictedData,
out Loss);
}
}
}
Debug.WriteLine($"{iter + 1}/{niterations}");
progressCallback?.Invoke($"{iter + 1}/{niterations}");
}
Debug.WriteLine("\nDone training!\n");
#endregion
}
return(HalvesForDenoising1.ToArray(), HalvesForDenoising2.ToArray(), StatsForDenoising.ToArray());
}