private static void extract_voxel(string lfile, string rfile, string prefix)
{
int w = 1920;
int h = 1080;
int shift = 0;
int count = 0;
VideoCapture lcap = new VideoCapture(lfile);
VideoCapture rcap = new VideoCapture(rfile);
while (true)
{
Image l = LoadArgs.get_image_from_stream(lcap);
Image r = LoadArgs.get_image_from_stream(rcap);
if (l.W == 0 || r.W == 0)
{
break;
}
if (count % 100 == 0)
{
shift = LoadArgs.best_3d_shift_r(l, r, -l.H / 100, l.H / 100);
Console.Write($"{shift}\n");
}
Image ls = LoadArgs.crop_image(l, (l.W - w) / 2, (l.H - h) / 2, w, h);
Image rs = LoadArgs.crop_image(r, 105 + (r.W - w) / 2, (r.H - h) / 2 + shift, w, h);
string buff = $"{prefix}_{count:05}_l";
LoadArgs.save_image(ls, buff);
buff = $"{prefix}_{count:05}_r";
LoadArgs.save_image(rs, buff);
++count;
}
}
private static void label_classifier(string datacfg, string filename, string weightfile)
{
int i;
Network net = Parser.parse_network_cfg(filename);
Network.set_batch_network(net, 1);
if (string.IsNullOrEmpty(weightfile))
{
Parser.load_weights(net, weightfile);
}
var options = OptionList.read_data_cfg(datacfg);
string labelList = OptionList.option_find_str(options, "names", "Data.Data/labels.list");
string testList = OptionList.option_find_str(options, "test", "Data.Data/train.list");
int classes = OptionList.option_find_int(options, "classes", 2);
string[] labels = Data.Data.get_labels(labelList);
string[] paths = Data.Data.GetPaths(testList);
int m = paths.Length;
for (i = 0; i < m; ++i)
{
Image im = LoadArgs.load_image_color(paths[i], 0, 0);
Image resized = LoadArgs.resize_min(im, net.W);
Image crop = LoadArgs.crop_image(resized, (resized.W - net.W) / 2, (resized.H - net.H) / 2, net.W, net.H);
float[] pred = Network.network_predict(net, crop.Data);
int ind = Utils.max_index(pred, classes);
Console.Write($"%s\n", labels[ind]);
}
}
private static void optimize_picture(Network net, Image orig, int maxLayer, float scale, float rate, float thresh, bool norm)
{
net.N = maxLayer + 1;
int dx = Utils.Rand.Next() % 16 - 8;
int dy = Utils.Rand.Next() % 16 - 8;
bool flip = Utils.Rand.Next() % 2 != 0;
Image crop = LoadArgs.crop_image(orig, dx, dy, orig.W, orig.H);
Image im = LoadArgs.resize_image(crop, (int)(orig.W * scale), (int)(orig.H * scale));
if (flip)
{
LoadArgs.flip_image(im);
}
Network.resize_network(net, im.W, im.H);
Layer last = net.Layers[net.N - 1];
Image delta = new Image(im.W, im.H, im.C);
NetworkState state = new NetworkState();
state.Input = (float[])im.Data.Clone();
state.Delta = (float[])im.Data.Clone();
Network.forward_network_gpu(net, state);
Blas.copy_ongpu(last.Outputs, last.OutputGpu, last.DeltaGpu);
Array.Copy(last.DeltaGpu, last.Delta, last.Outputs);
calculate_loss(last.Delta, last.Delta, last.Outputs, thresh);
Array.Copy(last.Delta, last.DeltaGpu, last.Outputs);
Network.backward_network_gpu(net, state);
Array.Copy(state.Delta, delta.Data, im.W * im.H * im.C);
if (flip)
{
LoadArgs.flip_image(delta);
}
Image resized = LoadArgs.resize_image(delta, orig.W, orig.H);
Image outi = LoadArgs.crop_image(resized, -dx, -dy, orig.W, orig.H);
if (norm)
{
Utils.normalize_array(outi.Data, outi.W * outi.H * outi.C);
}
Blas.Axpy_cpu(orig.W * orig.H * orig.C, rate, outi.Data, orig.Data);
LoadArgs.constrain_image(orig);
}
private static void try_classifier(string datacfg, string cfgfile, string weightfile, string filename, int layerNum)
{
Network net = Parser.parse_network_cfg(cfgfile);
if (string.IsNullOrEmpty(weightfile))
{
Parser.load_weights(net, weightfile);
}
Network.set_batch_network(net, 1);
Utils.Rand = new Random(2222222);
var options = OptionList.read_data_cfg(datacfg);
string nameList = OptionList.option_find_str(options, "names", "");
if (string.IsNullOrEmpty(nameList))
{
nameList = OptionList.option_find_str(options, "labels", "Data.Data/labels.list");
}
int top = OptionList.option_find_int(options, "top", 1);
int i = 0;
string[] names = Data.Data.get_labels(nameList);
var sw = new Stopwatch();
int[] indexes = new int[top];
string input = "";
while (true)
{
if (!string.IsNullOrEmpty(filename))
{
input = filename;
}
else
{
Console.Write($"Enter Image Path: ");
input = Console.ReadLine();
if (string.IsNullOrEmpty(input))
{
return;
}
input = input.TrimEnd();
}
Image orig = LoadArgs.load_image_color(input, 0, 0);
Image r = LoadArgs.resize_min(orig, 256);
Image im = LoadArgs.crop_image(r, (r.W - 224 - 1) / 2 + 1, (r.H - 224 - 1) / 2 + 1, 224, 224);
float[] mean = { 0.48263312050943f, 0.45230225481413f, 0.40099074308742f };
float[] std = { 0.22590347483426f, 0.22120921437787f, 0.22103996251583f };
float[] var = new float[3];
var[0] = std[0] * std[0];
var[1] = std[1] * std[1];
var[2] = std[2] * std[2];
Blas.Normalize_cpu(im.Data, mean, var, 1, 3, im.W * im.H);
float[] x = im.Data;
sw.Reset();
sw.Start();
float[] predictions = Network.network_predict(net, x);
Layer l = net.Layers[layerNum];
for (i = 0; i < l.C; ++i)
{
if (l.RollingMean.Length > i)
{
Console.Write($"%f %f %f\n", l.RollingMean[i], l.RollingVariance[i], l.Scales[i]);
}
}
Array.Copy(l.OutputGpu, l.Output, l.Outputs);
for (i = 0; i < l.Outputs; ++i)
{
Console.Write($"%f\n", l.Output[i]);
}
Network.top_predictions(net, top, indexes);
sw.Stop();
Console.Write($"%s: Predicted ini %f seconds.\n", input, sw.Elapsed.Seconds);
for (i = 0; i < top; ++i)
{
int index = indexes[i];
Console.Write($"%s: %f\n", names[index], predictions[index]);
}
if (!string.IsNullOrEmpty(filename))
{
break;
}
}
}
private static void validate_classifier_single(string datacfg, string filename, string weightfile)
{
int i, j;
Network net = Parser.parse_network_cfg(filename);
if (string.IsNullOrEmpty(weightfile))
{
Parser.load_weights(net, weightfile);
}
Network.set_batch_network(net, 1);
var options = OptionList.read_data_cfg(datacfg);
string labelList = OptionList.option_find_str(options, "labels", "Data.Data/labels.list");
string leafList = OptionList.option_find_str(options, "leaves", "");
if (!string.IsNullOrEmpty(leafList))
{
net.Hierarchy.Change_leaves(leafList);
}
string validList = OptionList.option_find_str(options, "valid", "Data.Data/train.list");
int classes = OptionList.option_find_int(options, "classes", 2);
int topk = OptionList.option_find_int(options, "top", 1);
string[] labels = Data.Data.get_labels(labelList);
string[] paths = Data.Data.GetPaths(validList);
int m = paths.Length;
float avgAcc = 0;
float avgTopk = 0;
int[] indexes = new int[topk];
for (i = 0; i < m; ++i)
{
int class2 = -1;
string path = paths[i];
for (j = 0; j < classes; ++j)
{
if (path.Contains(labels[j]))
{
class2 = j;
break;
}
}
Image im = LoadArgs.load_image_color(paths[i], 0, 0);
Image resized = LoadArgs.resize_min(im, net.W);
Image crop = LoadArgs.crop_image(resized, (resized.W - net.W) / 2, (resized.H - net.H) / 2, net.W, net.H);
float[] pred = Network.network_predict(net, crop.Data);
if (net.Hierarchy != null)
{
net.Hierarchy.Hierarchy_predictions(pred, 0, net.Outputs, false);
}
Utils.top_k(pred, classes, topk, indexes);
if (indexes[0] == class2)
{
avgAcc += 1;
}
for (j = 0; j < topk; ++j)
{
if (indexes[j] == class2)
{
avgTopk += 1;
}
}
Console.Write($"%d: top 1: %f, top %d: %f\n", i, avgAcc / (i + 1), topk, avgTopk / (i + 1));
}
}
private static void validate_classifier_10(string datacfg, string filename, string weightfile)
{
int i, j;
Network net = Parser.parse_network_cfg(filename);
Network.set_batch_network(net, 1);
if (string.IsNullOrEmpty(weightfile))
{
Parser.load_weights(net, weightfile);
}
var options = OptionList.read_data_cfg(datacfg);
string labelList = OptionList.option_find_str(options, "labels", "Data.Data/labels.list");
string validList = OptionList.option_find_str(options, "valid", "Data.Data/train.list");
int classes = OptionList.option_find_int(options, "classes", 2);
int topk = OptionList.option_find_int(options, "top", 1);
string[] labels = Data.Data.get_labels(labelList);
string[] paths = Data.Data.GetPaths(validList);
int m = paths.Length;
float avgAcc = 0;
float avgTopk = 0;
int[] indexes = new int[topk];
for (i = 0; i < m; ++i)
{
int class2 = -1;
string path = paths[i];
for (j = 0; j < classes; ++j)
{
if (path.Contains(labels[j]))
{
class2 = j;
break;
}
}
int w = net.W;
int h = net.H;
int shift = 32;
Image im = LoadArgs.load_image_color(paths[i], w + shift, h + shift);
Image[] images = new Image[10];
images[0] = LoadArgs.crop_image(im, -shift, -shift, w, h);
images[1] = LoadArgs.crop_image(im, shift, -shift, w, h);
images[2] = LoadArgs.crop_image(im, 0, 0, w, h);
images[3] = LoadArgs.crop_image(im, -shift, shift, w, h);
images[4] = LoadArgs.crop_image(im, shift, shift, w, h);
LoadArgs.flip_image(im);
images[5] = LoadArgs.crop_image(im, -shift, -shift, w, h);
images[6] = LoadArgs.crop_image(im, shift, -shift, w, h);
images[7] = LoadArgs.crop_image(im, 0, 0, w, h);
images[8] = LoadArgs.crop_image(im, -shift, shift, w, h);
images[9] = LoadArgs.crop_image(im, shift, shift, w, h);
float[] pred = new float[classes];
for (j = 0; j < 10; ++j)
{
float[] p = Network.network_predict(net, images[j].Data);
if (net.Hierarchy != null)
{
net.Hierarchy.Hierarchy_predictions(p, 0, net.Outputs, true);
}
Blas.Axpy_cpu(classes, 1, p, pred);
}
Utils.top_k(pred, classes, topk, indexes);
if (indexes[0] == class2)
{
avgAcc += 1;
}
for (j = 0; j < topk; ++j)
{
if (indexes[j] == class2)
{
avgTopk += 1;
}
}
Console.Write($"%d: top 1: %f, top %d: %f\n", i, avgAcc / (i + 1), topk, avgTopk / (i + 1));
}
}
public static void run_nightmare(List <string> args)
{
if (args.Count < 4)
{
Console.Error.Write($"usage: %s %s [cfg] [weights] [Image] [Layer] [options! (optional)]\n", args[0], args[1]);
return;
}
string cfg = args[2];
string weights = args[3];
string input = args[4];
int maxLayer = int.Parse(args[5]);
int range = Utils.find_int_arg(args, "-range", 1);
bool norm = Utils.find_int_arg(args, "-norm", 1) != 0;
int rounds = Utils.find_int_arg(args, "-rounds", 1);
int iters = Utils.find_int_arg(args, "-iters", 10);
int octaves = Utils.find_int_arg(args, "-octaves", 4);
float zoom = Utils.find_int_arg(args, "-zoom", 1);
float rate = Utils.find_int_arg(args, "-rate", .04f);
float thresh = Utils.find_int_arg(args, "-thresh", 1);
float rotate = Utils.find_int_arg(args, "-rotate", 0);
float momentum = Utils.find_int_arg(args, "-momentum", .9f);
float lambda = Utils.find_int_arg(args, "-lambda", .01f);
string prefix = Utils.find_int_arg(args, "-prefix", "");
bool reconstruct = Utils.find_arg(args, "-reconstruct");
int smoothSize = Utils.find_int_arg(args, "-smooth", 1);
Network net = Parser.parse_network_cfg(cfg);
Parser.load_weights(net, weights);
string cfgbase = Utils.Basecfg(cfg);
string imbase = Utils.Basecfg(input);
Network.set_batch_network(net, 1);
Image im = LoadArgs.load_image_color(input, 0, 0);
float[] features = new float[0];
Image update = null;
if (reconstruct)
{
Network.resize_network(net, im.W, im.H);
int zz = 0;
Network.network_predict(net, im.Data);
Image outIm = Network.get_network_image(net);
Image crop = LoadArgs.crop_image(outIm, zz, zz, outIm.W - 2 * zz, outIm.H - 2 * zz);
Image fIm = LoadArgs.resize_image(crop, outIm.W, outIm.H);
Console.Write($"%d features\n", outIm.W * outIm.H * outIm.C);
im = LoadArgs.resize_image(im, im.W, im.H);
fIm = LoadArgs.resize_image(fIm, fIm.W, fIm.H);
features = fIm.Data;
int i;
for (i = 0; i < 14 * 14 * 512; ++i)
{
features[i] += Utils.rand_uniform(-.19f, .19f);
}
im = LoadArgs.make_random_image(im.W, im.H, im.C);
update = new Image(im.W, im.H, im.C);
}
int e;
int n;
for (e = 0; e < rounds; ++e)
{
Console.Error.Write($"Iteration: ");
for (n = 0; n < iters; ++n)
{
Console.Error.Write($"%d, ", n);
if (reconstruct)
{
reconstruct_picture(net, features, im, update, rate, momentum, lambda, smoothSize, 1);
//if ((n+1)%30 == 0) rate *= .5;
LoadArgs.show_image(im, "reconstruction");
CvInvoke.WaitKey(10);
}
else
{
int layer = maxLayer + Utils.Rand.Next() % range - range / 2;
int octave = Utils.Rand.Next() % octaves;
optimize_picture(net, im, layer, 1 / (float)Math.Pow(1.33333333, octave), rate, thresh, norm);
}
}
Console.Error.Write($"done\n");
string buff;
if (!string.IsNullOrEmpty(prefix))
{
buff = $"{prefix}_{imbase}_{cfgbase}_{maxLayer}_{e:06}%s/%s_%s_%d_%06d";
}
else
{
buff = $"{imbase}_{cfgbase}_{maxLayer}_{e:06}";
}
Console.Write($"%d %s\n", e, buff);
LoadArgs.save_image(im, buff);
//LoadArgs.show_image(im, buff);
//CvInvoke.WaitKey();
if (rotate != 0)
{
Image rot = LoadArgs.rotate_image(im, rotate);
im = rot;
}
Image crop = LoadArgs.crop_image(im, (int)(im.W * (1f - zoom) / 2f), (int)(im.H * (1f - zoom) / 2f), (int)(im.W * zoom), (int)(im.H * zoom));
Image resized = LoadArgs.resize_image(crop, im.W, im.H);
im = resized;
}
}
