Update Figures: naming and documentation

Author: schilling40

flamingo_tools/segmentation/cochlea_mapping.py

@@ -557,7 +557,7 @@ def map_frequency(table: pd.DataFrame, animal: str = "mouse", otof: bool = False
     if otof and animal == "mouse":
         # freq_min = 4.84 kHz
         # freq_max = 78.8 kHz
-        # Mueller, Hearing Research 202 (2005) 63–73, https://doi.org/10.1016/j.heares.2004.08.011
+        # Mueller, Hearing Research 202 (2005) 63-73, https://doi.org/10.1016/j.heares.2004.08.011
         # function has format f(x) = 10 ** (a * (k - (1-x)))
         var_a = 100 / 82.5
         var_k = 1.565

reproducibility/tonotopic_mapping/2025-11-IHC_PELCO.json renamed to reproducibility/tonotopic_mapping/fHC_IHC.json

@@ -4,6 +4,8 @@
 import imageio.v3 as imageio
 import numpy as np
 import pandas as pd
+from glob import glob
+from pathlib import Path
 
 from flamingo_tools.validation import compute_matches_for_annotated_slice
 
@@ -291,12 +293,69 @@ def print_accuracy_ihc():
         print_accuracy(os.path.join(seg_dir, baseline))
 
 
+def runtimes_sgn():
+    for_comparison = ["distance_unet", "micro-sam", "cellpose3", "cellpose-sam", "stardist"]
+
+    cochlea_dir = "/mnt/vast-nhr/projects/nim00007/data/moser/cochlea-lightsheet"
+    val_sgn_dir = f"{cochlea_dir}/predictions/val_sgn"
+    image_dir = f"{cochlea_dir}/AnnotatedImageCrops/F1ValidationSGNs/for_consensus_annotation"
+
+    image_paths = sorted(glob(os.path.join(image_dir, "*.tif")))
+
+    runtimes = {name: [] for name in for_comparison}
+
+    for path in image_paths:
+        eval_fname = Path(path).stem + "_dic.json"
+        for seg_name in for_comparison:
+            eval_path = os.path.join(val_sgn_dir, seg_name, eval_fname)
+            with open(eval_path, "r") as f:
+                result = json.load(f)
+            rt = result["time"]
+            runtimes[seg_name].append(rt)
+
+    for name, rts in runtimes.items():
+        print(name, ":", np.mean(rts), "+-", np.std(rts))
+
+
+def runtimes_ihc():
+    for_comparison = ["distance_unet_v3", "micro-sam", "cellpose3", "cellpose-sam"]
+
+    cochlea_dir = "/mnt/vast-nhr/projects/nim00007/data/moser/cochlea-lightsheet"
+    val_sgn_dir = f"{cochlea_dir}/predictions/val_ihc"
+    image_dir = f"{cochlea_dir}/AnnotatedImageCrops/F1ValidationIHCs"
+
+    image_paths = sorted(glob(os.path.join(image_dir, "*.tif")))
+
+    runtimes = {name: [] for name in for_comparison}
+
+    for path in image_paths:
+        eval_fname = Path(path).stem + "_dic.json"
+        for seg_name in for_comparison:
+            eval_path = os.path.join(val_sgn_dir, seg_name, eval_fname)
+            if not os.path.exists(eval_path):
+                continue
+            with open(eval_path, "r") as f:
+                result = json.load(f)
+            rt = result["time"]
+            runtimes[seg_name].append(rt)
+
+    for name, rts in runtimes.items():
+        print(name, ":", np.mean(rts), "+-", np.std(rts))
+
+
 def main():
     eval_all_sgn()
     eval_all_ihc()
     print_accuracy_sgn()
     print_accuracy_ihc()
 
+    # average runtimes and standard deviation
+    print("SGNs:")
+    runtimes_sgn()
+    print()
+    print("IHCs:")
+    runtimes_ihc()
+
 
 if __name__ == "__main__":
     main()

scripts/baselines/eval_baseline.py

@@ -21,206 +21,11 @@
 COLOR_LITERATURE = "#27339C"
 
 
-def plot_legend_suppfig02(save_path):
-    """Plot common legend for figure 2c.
-
-    Args:
-        save_path: save path to save legend.
-    """
-    # Colors
-    color = [COLOR_P, COLOR_R, COLOR_F, COLOR_T]
-    label = ["Precision", "Recall", "F1-score", "Processing time"]
-
-    handles = [get_flatline_handle(c) for c in color]
-    legend = plt.legend(handles, label, loc=3, ncol=len(label), framealpha=1, frameon=False)
-    export_legend(legend, save_path)
-    legend.remove()
-    plt.close()
-
-
-def supp_fig_02(save_path, plot=False, segm="SGN", mode="precision"):
-    # SGN
-    value_dict = {
-        "SGN": {
-            "stardist": {
-                "label": "Stardist",
-                "precision": 0.706,
-                "recall": 0.630,
-                "f1-score": 0.628,
-                "marker": "*",
-                "runtime": 536.5,
-                "runtime_std": 148.4
-            },
-            "micro_sam": {
-                "label": "µSAM",
-                "precision": 0.140,
-                "recall": 0.782,
-                "f1-score": 0.228,
-                "marker": "D",
-                "runtime": 407.5,
-                "runtime_std": 107.5
-            },
-            "cellpose_3": {
-                "label": "Cellpose 3",
-                "precision": 0.117,
-                "recall": 0.607,
-                "f1-score": 0.186,
-                "marker": "v",
-                "runtime": 167.9116359,
-                "runtime_std": 40.2,
-            },
-            "cellpose_sam": {
-                "label": "Cellpose-SAM",
-                "precision": 0.250,
-                "recall": 0.003,
-                "f1-score": 0.005,
-                "marker": "^",
-                "runtime": 2232.007748,
-                "runtime_std": None,
-            },
-            "spiner2D": {
-                "label": "Spiner",
-                "precision": 0.373,
-                "recall": 0.340,
-                "f1-score": 0.326,
-                "marker": "o",
-                "runtime": None,
-                "runtime_std": None,
-            },
-            "distance_unet": {
-                "label": "CochleaNet",
-                "precision": 0.886,
-                "recall": 0.804,
-                "f1-score": 0.837,
-                "marker": "s",
-                "runtime": 168.8,
-                "runtime_std": 21.8
-            },
-        },
-        "IHC": {
-            "micro_sam": {
-                "label": "µSAM",
-                "precision": 0.053,
-                "recall": 0.684,
-                "f1-score": 0.094,
-                "marker": "D",
-                "runtime": 445.6,
-                "runtime_std": 106.6
-            },
-            "cellpose_3": {
-                "label": "Cellpose 3",
-                "precision": 0.375,
-                "recall": 0.554,
-                "f1-score": 0.329,
-                "marker": "v",
-                "runtime": 162.3493934,
-                "runtime_std": 30.1,
-            },
-            "cellpose_sam": {
-                "label": "Cellpose-SAM",
-                "precision": 0.636,
-                "recall": 0.025,
-                "f1-score": 0.047,
-                "marker": "^",
-                "runtime": 2137.944779,
-                "runtime_std": None
-            },
-            "distance_unet": {
-                "label": "CochleaNet",
-                "precision": 0.693,
-                "recall": 	0.567,
-                "f1-score": 0.618,
-                "marker": "s",
-                "runtime": 69.01,
-                "runtime_std": None
-            },
-        }
-    }
-
-    # Convert setting labels to numerical x positions
-    offset = 0.08  # horizontal shift for scatter separation
-
-    # Plot
-    tick_rotation = 0
-
-    main_label_size = 20
-    main_tick_size = 16
-    marker_size = 200
-
-    labels = [value_dict[segm][key]["label"] for key in value_dict[segm].keys()]
-
-    if mode == "precision":
-        fig, ax = plt.subplots(figsize=(10, 5))
-        # Convert setting labels to numerical x positions
-        offset = 0.08  # horizontal shift for scatter separation
-        for num, key in enumerate(list(value_dict[segm].keys())):
-            precision = [value_dict[segm][key]["precision"]]
-            recall = [value_dict[segm][key]["recall"]]
-            f1score = [value_dict[segm][key]["f1-score"]]
-            marker = value_dict[segm][key]["marker"]
-            x_pos = num + 1
-
-            plt.scatter([x_pos - offset], precision, label="Precision manual",
-                        color=COLOR_P, marker=marker, s=marker_size)
-            plt.scatter([x_pos],         recall, label="Recall manual",
-                        color=COLOR_R, marker=marker, s=marker_size)
-            plt.scatter([x_pos + offset], f1score, label="F1-score manual",
-                        color=COLOR_F, marker=marker, s=marker_size)
-
-        # Labels and formatting
-        x_pos = np.arange(1, len(labels)+1)
-        plt.xticks(x_pos, labels, fontsize=main_tick_size, rotation=tick_rotation)
-        plt.yticks(fontsize=main_tick_size)
-        plt.ylabel("Value", fontsize=main_label_size)
-        plt.ylim(-0.1, 1)
-        # plt.legend(loc="lower right", fontsize=legendsize)
-        plt.grid(axis="y", linestyle="solid", alpha=0.5)
-
-    elif mode == "runtime":
-        fig, ax = plt.subplots(figsize=(8.5, 5))
-        if "Spiner" in labels:
-            labels.remove("Spiner")
-
-        # Convert setting labels to numerical x positions
-        offset = 0.08  # horizontal shift for scatter separation
-        x_pos = 1
-        for num, key in enumerate(list(value_dict[segm].keys())):
-            runtime = [value_dict[segm][key]["runtime"]]
-            if runtime[0] is None:
-                continue
-            marker = value_dict[segm][key]["marker"]
-            plt.scatter([x_pos], runtime, label="Runtime", color=COLOR_T, marker=marker, s=marker_size)
-            x_pos = x_pos + 1
-
-        # Labels and formatting
-        x_pos = np.arange(1, len(labels)+1)
-        plt.xticks(x_pos, labels, fontsize=16, rotation=tick_rotation)
-        plt.yticks(fontsize=main_tick_size)
-        plt.ylabel("Processing time [s]", fontsize=main_label_size)
-        plt.ylim(10, 2600)
-        plt.yscale('log')
-        # plt.legend(loc="lower right", fontsize=legendsize)
-        plt.grid(axis="y", linestyle="solid", alpha=0.5)
-
-    else:
-        raise ValueError("Unsupported mode for plotting.")
-
-    plt.tight_layout()
-    prism_cleanup_axes(ax)
-
-    if ".png" in save_path:
-        plt.savefig(save_path, bbox_inches="tight", pad_inches=0.1, dpi=png_dpi)
-    else:
-        plt.savefig(save_path, bbox_inches='tight', pad_inches=0)
-
-    if plot:
-        plt.show()
-    else:
-        plt.close()
-
-
-def plot_legend_fig02c(save_path, plot_mode="shapes"):
-    """Plot common legend for figure 2c.
+def plot_legend_fig02c(
+    save_path: str,
+    plot_mode: str = "shapes",
+):
+    """Plot common legend for Figure 2c.
 
     Args:.
         save_path: save path to save legend.
@@ -253,7 +58,10 @@ def plot_legend_fig02c(save_path, plot_mode="shapes"):
         raise ValueError("Choose either 'shapes' or 'colors' as plot_mode.")
 
 
-def fig_02c(save_path, plot=False, all_versions=False):
+def fig_02c(
+    save_path: str,
+    plot: bool = False,
+):
     """Scatter plot showing the precision, recall, and F1-score of SGN (distance U-Net, manual),
     IHC (distance U-Net, manual), and synapse detection (U-Net).
     """
@@ -336,7 +144,11 @@ def _load_ribbon_synapse_counts():
     return syn_counts
 
 
-def fig_02d(save_path, plot=False, plot_average_ribbon_synapses=False):
+def fig_02d(
+    save_path: str,
+    plot: bool = False,
+    plot_average_ribbon_synapses: bool = False,
+):
     """Box plot showing the counts for SGN and IHC per (mouse) cochlea in comparison to literature values.
     """
     prism_style()
@@ -446,31 +258,22 @@ def fig_02d(save_path, plot=False, plot_average_ribbon_synapses=False):
 
 
 def main():
-    parser = argparse.ArgumentParser(description="Generate plots for Fig 2 of the cochlea paper.")
+    parser = argparse.ArgumentParser(description="Generate plots for Figure 2 of the CochleaNet paper.")
     parser.add_argument("--figure_dir", "-f", type=str, help="Output directory for plots.", default="./panels/fig2")
     parser.add_argument("--plot", action="store_true")
     args = parser.parse_args()
 
     os.makedirs(args.figure_dir, exist_ok=True)
 
     # Panel C: Evaluation of the segmentation results:
-    fig_02c(save_path=os.path.join(args.figure_dir, f"fig_02c.{FILE_EXTENSION}"), plot=args.plot, all_versions=False)
+    fig_02c(save_path=os.path.join(args.figure_dir, f"fig_02c.{FILE_EXTENSION}"), plot=args.plot)
     plot_legend_fig02c(os.path.join(args.figure_dir, f"fig_02c_legend_shapes.{FILE_EXTENSION}"), plot_mode="shapes")
     plot_legend_fig02c(os.path.join(args.figure_dir, f"fig_02c_legend_colors.{FILE_EXTENSION}"), plot_mode="colors")
 
     # Panel D: The number of SGNs, IHCs and average number of ribbon synapses per IHC
     fig_02d(save_path=os.path.join(args.figure_dir, f"fig_02d.{FILE_EXTENSION}"),
             plot=args.plot, plot_average_ribbon_synapses=True)
 
-    # Supplementary Figure 2: Comparing other methods in terms of segmentation accuracy and runtime
-    plot_legend_suppfig02(save_path=os.path.join(args.figure_dir, f"figsupp_02_legend_colors.{FILE_EXTENSION}"))
-    supp_fig_02(save_path=os.path.join(args.figure_dir, f"figsupp_02_sgn.{FILE_EXTENSION}"), segm="SGN")
-    supp_fig_02(save_path=os.path.join(args.figure_dir, f"figsupp_02_ihc.{FILE_EXTENSION}"), segm="IHC")
-    supp_fig_02(save_path=os.path.join(args.figure_dir, f"figsupp_02_sgn_time.{FILE_EXTENSION}"),
-                segm="SGN", mode="runtime")
-    supp_fig_02(save_path=os.path.join(args.figure_dir, f"figsupp_02_ihc_time.{FILE_EXTENSION}"),
-                segm="IHC", mode="runtime")
-
 
 if __name__ == "__main__":
     main()