Merge pull request #140 from radionets-project/symmetry_training

Utilize symmetry

.gitignore

@@ -44,3 +44,4 @@ __pycache__/
 dist/
 radionets/dl_framework/values.yaml
 *.csv
+*.swp

CHANGES.rst

@@ -1,3 +1,38 @@
+Radionets 0.2.0 (2023-01-31)
+============================
+
+
+API Changes
+-----------
+
+- train on half-sized iamges and applying symmetry afterward is a backward incompatible change
+  models trained with early versions of `radionets` are not supported anymore [`#140 <https://github.com/radionets-project/radionets/pull/140>`__]
+
+
+Bug Fixes
+---------
+
+- fixed sampling of test data set
+  fixed same indices for plots [`#140 <https://github.com/radionets-project/radionets/pull/140>`__]
+
+
+New Features
+------------
+
+- enabled training and evaluation of half sized images (for 128 pixel images) [`#140 <https://github.com/radionets-project/radionets/pull/140>`__]
+
+
+Maintenance
+-----------
+
+- Deleted unusable functions for new source types
+  Deleted unused hardcoded scaling [`#140 <https://github.com/radionets-project/radionets/pull/140>`__]
+
+
+Refactoring and Optimization
+----------------------------
+
+
 Radionets 0.1.18 (2023-01-30)
 =============================
 

docs/changes/140.bugfixes.rst

@@ -0,0 +1,2 @@
+fixed sampling of test data set
+fixed same indices for plots

docs/changes/140.feature.rst

@@ -0,0 +1 @@
+enabled training and evaluation of half sized images (for 128 pixel images)

docs/changes/140.maintenance.rst

@@ -0,0 +1,2 @@
+Deleted unusable functions for new source types
+Deleted unused hardcoded scaling

examples/default_eval_config.toml

@@ -24,7 +24,6 @@ visualize_prediction = true
 visualize_source_reconstruction = true
 visualize_contour = true
 visualize_dynamic_range = false
-visualize_blobs = true
 visualize_ms_ssim = false
 random = false
 num_images = 5

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "radionets"
-version = "0.1.16"
+version = "0.2.0"
 authors = [
   { name="Kevin Schmidt", email="[email protected]" },
 ]

radionets/dl_framework/architectures/res_exp.py

@@ -151,9 +151,7 @@ def __init__(self):
             nn.Conv2d(64, 64, 3, stride=1, padding=1, bias=False), nn.BatchNorm2d(64)
         )
 
-        self.final = nn.Sequential(
-            nn.Conv2d(64, 2, 9, stride=1, padding=4, groups=2),
-        )
+        self.final = nn.Sequential(nn.Conv2d(64, 2, 9, stride=1, padding=4, groups=2))
 
         self.hardtanh = nn.Hardtanh(-pi, pi)
         self.relu = nn.ReLU()
@@ -167,9 +165,9 @@ def forward(self, x):
 
         x = self.final(x)
 
-        x0 = x[:, 0].reshape(-1, 1, s, s)
+        x0 = x[:, 0].reshape(-1, 1, s // 2 + 1, s)
         x0 = self.relu(x0)
-        x1 = self.hardtanh(x[:, 1]).reshape(-1, 1, s, s)
+        x1 = self.hardtanh(x[:, 1]).reshape(-1, 1, s // 2 + 1, s)
 
         return torch.cat([x0, x1], dim=1)
 

radionets/dl_framework/architectures/unc_archs.py

@@ -4,36 +4,36 @@
     GeneralELU,
     LocallyConnected2d,
 )
-from radionets.dl_framework.architectures.res_exp import SRResNet
+from radionets.dl_framework.architectures.res_exp import SRResNet_16
 
 
 class Uncertainty(nn.Module):
     def __init__(self, img_size):
         super().__init__()
 
         self.conv1 = nn.Sequential(
-            nn.Conv2d(4, 8, 9, stride=1, padding=4, groups=2),
-            nn.BatchNorm2d(8),
+            nn.Conv2d(4, 16, 9, stride=1, padding=4, groups=2),
+            nn.BatchNorm2d(16),
             nn.PReLU(),
         )
 
         self.conv2 = nn.Sequential(
-            nn.Conv2d(8, 16, 3, stride=1, padding=1),
-            nn.BatchNorm2d(16),
+            nn.Conv2d(16, 32, 3, stride=1, padding=1),
+            nn.BatchNorm2d(32),
             nn.PReLU(),
         )
 
         self.conv3 = nn.Sequential(
-            nn.Conv2d(16, 32, 9, stride=1, padding=4, groups=2),
-            nn.BatchNorm2d(32),
+            nn.Conv2d(32, 64, 9, stride=1, padding=4, groups=2),
+            nn.BatchNorm2d(64),
             nn.PReLU(),
         )
 
         self.final = nn.Sequential(
             LocallyConnected2d(
-                32,
+                64,
                 2,
-                img_size,
+                [img_size // 2 + 1, img_size],
                 1,
                 stride=1,
                 bias=False,
@@ -55,7 +55,7 @@ def forward(self, x):
 class UncertaintyWrapper(nn.Module):
     def __init__(self, img_size):
         super().__init__()
-        self.pred = SRResNet()
+        self.pred = SRResNet_16()
 
         self.uncertainty = Uncertainty(img_size)
 

radionets/dl_framework/callbacks.py

@@ -183,7 +183,7 @@ class DataAug(Callback):
     def before_batch(self):
         x = self.xb[0].clone()
         y = self.yb[0].clone()
-        randint = np.random.randint(0, 4, x.shape[0])
+        randint = np.random.randint(0, 1, x.shape[0]) * 2
         last_axis = len(x.shape) - 1
         for i in range(x.shape[0]):
             x[i] = torch.rot90(x[i], int(randint[i]), [last_axis - 2, last_axis - 1])

radionets/dl_framework/data.py

@@ -71,6 +71,8 @@ def open_image(self, var, i):
         if data.shape[0] == 1:
             data = data.squeeze(0)
 
+        data = data[:, :65, :]
+
         return data.float()
 
 

radionets/dl_framework/model.py

@@ -191,7 +191,6 @@ def __init__(
         self, in_channels, out_channels, output_size, kernel_size, stride, bias=False
     ):
         super(LocallyConnected2d, self).__init__()
-        output_size = _pair(output_size)
         self.weight = nn.Parameter(
             torch.randn(
                 1,
@@ -281,10 +280,14 @@ def forward(self, x):
 
     def _conv_block(self, ni, nf, stride):
         return nn.Sequential(
-            nn.Conv2d(ni, nf, 3, stride=stride, padding=1, bias=False),
+            nn.Conv2d(
+                ni, nf, 3, stride=stride, padding=1, bias=False, padding_mode="reflect"
+            ),
             nn.BatchNorm2d(nf),
             nn.PReLU(),
-            nn.Conv2d(nf, nf, 3, stride=1, padding=1, bias=False),
+            nn.Conv2d(
+                nf, nf, 3, stride=1, padding=1, bias=False, padding_mode="reflect"
+            ),
             nn.BatchNorm2d(nf),
         )
 

radionets/evaluation/plotting.py

@@ -6,14 +6,13 @@
 import torch
 import matplotlib as mpl
 from matplotlib.colors import ListedColormap, LogNorm
-from matplotlib.lines import Line2D
-from matplotlib.patches import Arc, Rectangle
+from matplotlib.patches import Rectangle
 from mpl_toolkits.axes_grid1 import make_axes_locatable
 from pytorch_msssim import ms_ssim
-from radionets.evaluation.blob_detection import calc_blobs
+
 from radionets.evaluation.contour import compute_area_ratio
 from radionets.evaluation.dynamic_range import calc_dr, get_boxsize
-from radionets.evaluation.jet_angle import calc_jet_angle
+
 from radionets.evaluation.utils import (
     check_vmin_vmax,
     make_axes_nice,
@@ -209,11 +208,6 @@ def visualize_with_fourier(
     real_pred, imag_pred = img_pred[0], img_pred[1]
     real_truth, imag_truth = img_truth[0], img_truth[1]
 
-    if amp_phase:
-        inp_real = 10 ** (10 * inp_real - 10) - 1e-10
-        real_pred = 10 ** (10 * real_pred - 10) - 1e-10
-        real_truth = 10 ** (10 * real_truth - 10) - 1e-10
-
     # plotting
     fig, ((ax1, ax2, ax3), (ax4, ax5, ax6)) = plt.subplots(
         2, 3, figsize=(16, 10), sharex=True, sharey=True
@@ -291,21 +285,17 @@ def visualize_with_fourier_diff(
     real_pred, imag_pred = img_pred[0], img_pred[1]
     real_truth, imag_truth = img_truth[0], img_truth[1]
 
-    if amp_phase:
-        real_pred = 10 ** (10 * real_pred - 10) - 1e-10
-        real_truth = 10 ** (10 * real_truth - 10) - 1e-10
-
     # plotting
     # plt.style.use('./paper_large_3_2.rc')
     fig, ((ax1, ax2, ax3), (ax4, ax5, ax6)) = plt.subplots(
         2, 3, figsize=(16, 10), sharex=True, sharey=True
     )
 
     if amp_phase:
-        im1 = ax1.imshow(real_pred, cmap="inferno", norm=LogNorm())
+        im1 = ax1.imshow(real_pred, cmap="inferno")
         make_axes_nice(fig, ax1, im1, r"Amplitude Prediction")
 
-        im2 = ax2.imshow(real_truth, cmap="inferno", norm=LogNorm())
+        im2 = ax2.imshow(real_truth, cmap="inferno")
         make_axes_nice(fig, ax2, im2, r"Amplitude Truth")
 
         a = check_vmin_vmax(real_pred - real_truth)
@@ -344,37 +334,16 @@ def visualize_source_reconstruction(
     out_path,
     i,
     dr=False,
-    blobs=False,
     msssim=False,
     plot_format="png",
 ):
-    m_truth, n_truth, alpha_truth = calc_jet_angle(ifft_truth)
-    m_pred, n_pred, alpha_pred = calc_jet_angle(ifft_pred)
-    x_space = torch.arange(0, 63, 1)
-
     # plt.style.use("./paper_large_3.rc")
     fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(16, 10), sharey=True)
 
     # Plot prediction
-    ax1.plot(x_space, m_pred * x_space + n_pred, "w--", alpha=0.5)
-    ax1.axvline(32, 0, 1, linestyle="--", color="white", alpha=0.5)
-
-    # create angle visualization
-    theta1 = min(0, -alpha_pred.numpy()[0])
-    theta2 = max(0, -alpha_pred.numpy()[0])
-    ax1.add_patch(Arc([32, 32], 50, 50, 90, theta1, theta2, color="white"))
-
     im1 = ax1.imshow(ifft_pred, vmax=ifft_truth.max(), cmap="inferno")
 
     # Plot truth
-    ax2.plot(x_space, m_truth * x_space + n_truth, "w--", alpha=0.5)
-    ax2.axvline(32, 0, 1, linestyle="--", color="white", alpha=0.5)
-
-    # create angle visualization
-    theta1 = min(0, -alpha_truth.numpy()[0])
-    theta2 = max(0, -alpha_truth.numpy()[0])
-    ax2.add_patch(Arc([32, 32], 50, 50, 90, theta1, theta2, color="white"))
-
     im2 = ax2.imshow(ifft_truth, cmap="inferno")
 
     a = check_vmin_vmax(ifft_pred - ifft_truth)
@@ -389,15 +358,6 @@ def visualize_source_reconstruction(
     ax2.set_xlabel(r"Pixels")
     ax3.set_xlabel(r"Pixels")
 
-    # ax1.tick_params(axis="both", labelsize=20)
-    # ax2.tick_params(axis="both", labelsize=20)
-    # ax3.tick_params(axis="both", labelsize=20)
-
-    if blobs:
-        blobs_pred, blobs_truth = calc_blobs(ifft_pred, ifft_truth)
-        plot_blobs(blobs_pred, ax1)
-        plot_blobs(blobs_truth, ax2)
-
     if dr:
         dr_truth, dr_pred, num_boxes, corners = calc_dr(
             ifft_truth[None, ...], ifft_pred[None, ...]
@@ -417,15 +377,6 @@ def visualize_source_reconstruction(
 
     outpath = str(out_path) + f"/fft_pred_{i}.{plot_format}"
 
-    line = Line2D(
-        [], [], linestyle="-", color="w", label=rf"$\alpha = {alpha_pred[0]:.2f}\,$deg"
-    )
-    line_truth = Line2D(
-        [], [], linestyle="-", color="w", label=rf"$\alpha = {alpha_truth[0]:.2f}\,$deg"
-    )
-
-    ax1.legend(loc="best", handles=[line])
-    ax2.legend(loc="best", handles=[line_truth])
     fig.tight_layout(pad=1)
     plt.savefig(outpath, bbox_inches="tight", pad_inches=0.05)
     plt.close("all")
@@ -472,19 +423,19 @@ def visualize_uncertainty(
         2, 2, sharey=True, sharex=True, figsize=(12, 10)
     )
 
-    im1 = ax1.imshow(true_phase)
+    im1 = ax1.imshow(true_phase, cmap=OrBu, vmin=-np.pi, vmax=np.pi)
 
-    im2 = ax2.imshow(pred_phase)
+    im2 = ax2.imshow(pred_phase, cmap=OrBu, vmin=-np.pi, vmax=np.pi)
 
     im3 = ax3.imshow(unc_phase)
 
     a = check_vmin_vmax(true_phase - pred_phase)
     im4 = ax4.imshow(true_phase - pred_phase, cmap=OrBu, vmin=-a, vmax=a)
 
-    make_axes_nice(fig, ax1, im1, r"Simulation")
-    make_axes_nice(fig, ax2, im2, r"Predicted $\mu$")
+    make_axes_nice(fig, ax1, im1, r"Simulation", phase=True)
+    make_axes_nice(fig, ax2, im2, r"Predicted $\mu$", phase=True)
     make_axes_nice(fig, ax3, im3, r"Predicted $\sigma^2$", unc=True)
-    make_axes_nice(fig, ax4, im4, r"Difference")
+    make_axes_nice(fig, ax4, im4, r"Difference", phase_diff=True)
 
     ax1.set_ylabel(r"pixels")
     ax3.set_ylabel(r"pixels")

radionets/evaluation/scripts/start_evaluation.py

@@ -38,6 +38,7 @@ def main(configuration_path):
     print(eval_conf, "\n")
 
     if eval_conf["sample_unc"]:
+        click.echo("Sampling test data set.\n")
         save_sampled(eval_conf)
 
     for entry in conf["inspection"]:
@@ -66,9 +67,6 @@ def main(configuration_path):
 
         click.echo(f"\nCreated {eval_conf['num_images']} test predictions.\n")
 
-    if eval_conf["vis_blobs"]:
-        click.echo("\nBlob visualization is enabled for source plots.\n")
-
     if eval_conf["vis_ms_ssim"]:
         click.echo("\nVisualization of ms ssim is enabled for source plots.\n")