Commit c61cd18c authored by Andrey Filippov's avatar Andrey Filippov
Browse files

Working on metrics for the current heuristic DSI

parent 0b346308

src/main/java/MLStats.java

+65 −20
Original line number Diff line number Diff line
@@ -52,7 +52,7 @@ public class MLStats { 
//	}

	public static boolean dsiHistogram(String dir) {

		Path path= Paths.get(dir);

		int disparity_bins =         400;

		int disparity_bins =        1000;

		int strength_bins =          100;



		double disparity_min_drop =  -0.1;
@@ -64,6 +64,10 @@ public class MLStats { 
		double strength_max_drop =    1.0; //

		double strength_max_clip =    0.9; //

		boolean normalize =           true;

		double  master_weight_power = 1.0;

		double  master_weight_floor = 0.08;

		double  disparity_outlier =   1.0;





		String mask = ".*-DSI_COMBO\\.tiff";
@@ -80,6 +84,9 @@ public class MLStats { 
		gd.addNumericField("Drop tiles with strength above",                 strength_max_drop,            3);

		gd.addNumericField("Clip high strength with",                        strength_max_clip,            3);

		gd.addCheckbox("Normalize histogram to average 1.0",                 normalize);

		gd.addNumericField("Master weight power (after floor)",              master_weight_power,          3);

		gd.addNumericField("Master weight floor",                            master_weight_floor,          3);

		gd.addNumericField("Ignore tiles with disparity difference higher",  disparity_outlier,            3);



		gd.showDialog ();

		if (gd.wasCanceled()) return false;
@@ -95,6 +102,9 @@ public class MLStats { 
		strength_max_drop =    gd.getNextNumber();

		strength_max_clip =    gd.getNextNumber();

		normalize =            gd.getNextBoolean();

		master_weight_power =  gd.getNextNumber();

		master_weight_floor =  gd.getNextNumber();

		disparity_outlier =    gd.getNextNumber();

		// get list of all files:

		System.out.println("File mask = "+mask);
@@ -116,7 +126,9 @@ public class MLStats { 
			e.printStackTrace();

		}

		int [][] hist = new int [disparity_bins][strength_bins];

		int [] slices = {TwoQuadCLT.DSI_DISPARITY_RIG,TwoQuadCLT.DSI_STRENGTH_RIG};

		int [] slices = {TwoQuadCLT.DSI_DISPARITY_RIG,TwoQuadCLT.DSI_STRENGTH_RIG, TwoQuadCLT.DSI_DISPARITY_MAIN,TwoQuadCLT.DSI_STRENGTH_MAIN};

		double [][][] ds_error = new double [disparity_bins][strength_bins][2];

		double disparity_outlier2 = disparity_outlier*disparity_outlier;

		double disparity_step = (disparity_max_clip - disparity_min_clip) / disparity_bins;

		double strength_step =  (strength_max_clip -  strength_min_clip)  / strength_bins;

		double disparity_offs = disparity_min_clip - disparity_step/2; // last and first bin that include clip will be 0.5 width
@@ -155,31 +167,56 @@ public class MLStats { 
//					int [][] hist = new int [disparity_bins][strength_bins];

					  hist[dbin][sbin]++;

					  nut++;



					  double dm = dsi_float[2][nTile];

					  double sm = dsi_float[3][nTile] - master_weight_floor;

					  double de2 = (dm - d);

					  de2 *= de2;

					  if ((de2 <= disparity_outlier2) && (sm > 0.0)) {

						  double w = 1.0;

						  if (master_weight_power > 0.0) {

							  w = sm;

							  if (master_weight_power != 1.0) {

								w = Math.pow(w, master_weight_power);

							  }

							  ds_error[dbin][sbin][0] += w* de2;

							  ds_error[dbin][sbin][1] += w;

						  }

					  }

				  }

			  }

			  System.out.println(p.getFileName()+": "+nut+" useful tiles counted");

			  total_tiles_used += nut;

		}

	   System.out.println("Total number of useful tiles: "+total_tiles_used);

		double [] hist_double = new double [disparity_bins*strength_bins];

		double [][] hist_double = new double [2][disparity_bins*strength_bins];

		double scale = 1.0;

		if (normalize) {

			scale *= (1.0* disparity_bins * strength_bins) / total_tiles_used;

		}

		for (int nTile = 0; nTile < hist_double.length; nTile++) {

//		  ds_error[dbin][sbin][0] += w* de2;

//		  ds_error[dbin][sbin][1] += w;



		for (int nTile = 0; nTile < hist_double[0].length; nTile++) {

			int dbin = nTile % disparity_bins;

			int sbin = nTile / disparity_bins;

			hist_double[nTile] = scale * hist[dbin][sbin];

			hist_double[0][nTile] = scale * hist[dbin][sbin];

			if (ds_error[dbin][sbin][1] > 0.0) {

				hist_double[1][nTile] = Math.sqrt(ds_error[dbin][sbin][0]/ds_error[dbin][sbin][1]);

			} else {

				hist_double[1][nTile] = Double.NaN;

			}

		}

//		  ImagePlus imp= makeArrays(pixels, width, height, title);

//		  if (imp!=null) imp.show();



//		  ImagePlus imp = (new showDoubleFloatArrays()).makeArrays(dsi,quadCLT_main.tp.getTilesX(), quadCLT_main.tp.getTilesY(),  title, DSI_SLICES);

		String [] titles = {"histogram", "disp_err"};

		ImagePlus imp = (new showDoubleFloatArrays()).makeArrays(

				hist_double,

				disparity_bins,

				strength_bins,

				"DSI_histogram");

				"DSI_histogram",

				titles

				);

		imp.setProperty("disparity_bins",  disparity_bins+"");

		imp.setProperty("comment_disparity_bins",  "Number of disparity bins");

		imp.setProperty("strength_bins",  strength_bins+"");
@@ -206,6 +243,14 @@ public class MLStats { 
		imp.setProperty("total_tiles_used",  total_tiles_used+"");

		imp.setProperty("comment_total_tiles_used",  "Total number of tiles used");



		imp.setProperty("master_weight_power",  master_weight_power+"");

		imp.setProperty("comment_master_weight_power",  "Master weight power (after floor)");



		imp.setProperty("master_weight_floor",  master_weight_floor+"");

		imp.setProperty("comment_master_weight_floor",  "Master weight floor");



		imp.setProperty("disparity_outlier",  disparity_outlier+"");

		imp.setProperty("comment_disparity_outlier",  "Ignore tiles with disparity difference higher");



		(new JP46_Reader_camera(false)).encodeProperiesToInfo(imp);

		imp.show();