peer reviews

scripts/Paper/pr_figs.py

@@ -29,11 +29,20 @@ def add_model(this_df, this_axs, i, p, y="tp_precision", marker="r"):
 
 def plot_pr(losses:pd.DataFrame, title="str"):
     plot_params = ['volfrac', 'snr', 'cnr', 'particle_size', 'brightness', 'r']
-    titles		= ['Density $\phi$', 'SNR', 'CNR', 'Size ($\mu m$)', '$f_\mu$ (0-255)', 'Radius (pxls)']
+    titles		= ['Density $\phi$', 'SNR', 'CNR', 'Size ($\mu m$)', r'$f_{\bar{x}}$ (0-255)', 'Radius (pxls)']
+    xticks = [
+        [0.1, 0.25, 0.4, 0.55],
+        [0, 1, 5, 10],
+        [0, 1, 5, 10],
+        [0.1, 0.5, 1],
+        [0, 125, 255],
+        [4, 9, 14],
+    ]
 
     fig,axs = plt.subplots(2,len(plot_params),  sharey=True)
     plt.tight_layout(pad=0)
 
+    # Plot precision row
     this_axs = axs[0,:].flatten()
     for i, p in enumerate(plot_params):
         this_df = losses[losses['type'].isin([p])]
@@ -49,8 +58,10 @@ def plot_pr(losses:pd.DataFrame, title="str"):
             this_axs[i].set_ylabel("Precision", fontsize='large')
             this_axs[i].set_yticks([0,0.25,0.5,0.75,1])
             this_axs[i].set_ylim(-0.1,1.1)
+        if i == 5: 
             this_axs[i].legend(["TP", "U-net"])
 
+    # Plot recall row
     this_axs = axs[1,:].flatten()
     for i, p in enumerate(plot_params):
         this_df = losses[losses['type'].isin([p])]
@@ -62,6 +73,7 @@ def plot_pr(losses:pd.DataFrame, title="str"):
         this_axs[i].scatter(x=p, 		y = 'tp_recall', data=this_df, color='black', marker='<', alpha=0.5)
         this_axs[i].scatter(x=p, 		y = 'recall', 	data=this_df, color='red', marker='>', alpha=0.5)
         this_axs[i].set_xlabel(titles[i], fontsize='large')
+        this_axs[i].set_xticks(xticks[i])
         if i == 0: 
             this_axs[i].set_ylabel("Recall", fontsize='large')
 

scripts/figs/ap_on_sim.py

@@ -1,4 +1,3 @@
-from sklearn.inspection import plot_partial_dependence
 from colloidoscope import DeepColloid
 from colloidoscope.hoomd_sim_positions import read_gsd, convert_hoomd_positions
 from colloidoscope.simulator import crop_positions_for_label
@@ -17,9 +16,10 @@
 
 
 if __name__ == '__main__':
-	dataset_path = '/mnt/scratch/ak18001/Colloids/'
+	# dataset_path = '/mnt/scratch/ak18001/Colloids/'
 	# dataset_path = '/home/ak18001/Data/HDD/Colloids'
 	# dataset_path = '/home/wahab/Data/HDD/Colloids'
+	dataset_path = '/home/wahab/data/Colloids'
 	# dataset_path = '/mnt/storage/home/ak18001/scratch/Colloids'
 	dc = DeepColloid(dataset_path)
 
@@ -44,7 +44,7 @@
 	pred_pos = true_pos
 
 	ap, precisions, recalls, thresholds = dc.average_precision(true_pos, pred_pos, diameters=diameters)
-	dc.plot_pr(ap, precisions, recalls, thresholds, name='True', tag='o-', color='black', axs=axs[0,0], title='True positions')
+	dc.plot_pr(ap, precisions, recalls, thresholds, name='True', tag='o-', color='black', axs=axs[0,0], title='A - True positions')
 	# ap, precisions, recalls, thresholds = dc.average_precision(true_pos, pred_pos, diameters=diameters)
 	# fig = dc.plot_pr(ap, precisions, recalls, thresholds, name='Prediction', tag='x-', color='gray', axs=axs[0,0])
 	x, y = dc.get_gr(pred_pos, 100, 100)
@@ -55,7 +55,7 @@
 	pred_pos = true_pos[:len(true_pos)//2]
 
 	ap, precisions, recalls, thresholds = dc.average_precision(true_pos, pred_pos, diameters=diameters)
-	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='red', axs=axs[0,1], title='Half missing')
+	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='red', axs=axs[0,1], title='B - Half missing')
 	x, y = dc.get_gr(pred_pos, 100, 100)
 	dc.plot_gr(x, y, diameter, label=f'Pred', color='red', axs=axs[1,1])
 
@@ -71,7 +71,7 @@
 	pred_pos = np.concatenate([pred_pos, anomalies], axis=0)
 
 	ap, precisions, recalls, thresholds = dc.average_precision(true_pos, pred_pos, diameters=diameters)
-	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='green', axs=axs[0,2], title='Half added')
+	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='green', axs=axs[0,2], title='C - Half added')
 	x, y = dc.get_gr(pred_pos, 100, 100)
 	dc.plot_gr(x, y, diameter, label=f'Pred', color='green', axs=axs[1,2])
 
@@ -86,7 +86,7 @@
 	pred_pos = np.concatenate([pred_pos, anomalies], axis=0)
 
 	ap, precisions, recalls, thresholds = dc.average_precision(true_pos, pred_pos, diameters=diameters)
-	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='blue', axs=axs[0,3], title='Half missing/added')
+	dc.plot_pr(ap, precisions, recalls, thresholds, name='Pred', tag='o-', color='blue', axs=axs[0,3], title='D - Half missing/added')
 	x, y = dc.get_gr(pred_pos, 100, 100)
 	dc.plot_gr(x, y, diameter, label=f'Pred', color='blue', axs=axs[1,3])