mv analysis scripts to analysis

Author: trangle1302

bestfitting_ap_to_f1.py analysis/bestfitting_ap_to_f1.pybestfitting_ap_to_f1.py renamed to analysis/bestfitting_ap_to_f1.py

@@ -0,0 +1,57 @@
+import os
+import pandas as pd
+import numpy as np
+
+
+def average_precisions(y_true, y_pred):
+    _, classes = y_true.shape
+    average_precisions = []
+
+    for index in range(classes):
+        row_indices_sorted = np.argsort(-y_pred[:, index])
+
+        y_true_cls = y_true[row_indices_sorted, index]
+        y_pred_cls = y_pred[row_indices_sorted, index]
+
+        tp = y_true_cls == 1
+        fp = y_true_cls == 0
+
+        fp = np.cumsum(fp)
+        tp = np.cumsum(tp)
+
+        npos = np.sum(y_true_cls)
+
+        rec = tp * 1.0 / npos
+
+        # avoid divide by zero in case the first detection matches a difficult
+        # ground truth
+        prec = tp * 1.0 / np.maximum((tp + fp), np.finfo(np.float64).eps)
+
+        mrec = np.concatenate(([0.0], rec, [1.0]))
+        mpre = np.concatenate(([0.0], prec, [0.0]))
+
+        # compute the precision envelope
+        for i in range(mpre.size - 1, 0, -1):
+            mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
+
+            # to calculate area under PR curve, look for points
+            # where X axis (recall) changes value
+            i = np.where(mrec[1:] != mrec[:-1])[0]
+
+            # and sum (\Delta recall) * prec
+            average_precisions.append(np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]))
+
+    return average_precisions
+
+def main(d):
+    files = [f for f in os.listdir(d) if not f.endswith("_submission.csv")]
+    df = pd.DataFrame()
+    for f in files:
+        df_ = pd.read_csv(os.path.join(d, f))
+        df = df.append(df_, ignore_index=True)
+
+    # df['GT cell label'] = [f.split("\n")[0].split(" ")[-1] for f in df['GT cell label']]
+
+if __name__ == "__main__":
+    d = "/home/trangle/HPA_SingleCellClassification/examples/bestfitting"
+    main(d)

analysis/calculate_AP.py

@@ -261,8 +261,9 @@ def prepare_meta_publicHPAv21():
     thres_vector = thres_classes.best_threshold
     il_labels = tmp[[l+'_y' for l in LABEL_ALIASE_LIST]].values
     sc_labels = tmp[[l+'_x' for l in LABEL_ALIASE_LIST]].values
-    sc_labels = np.array([sc_labels.transpose()[c] > thres_vector[c] for c in range(19)]).transpose()
-    #sc_labels = il_labels*sc_labels 
+    sc_labels = np.array([sc_labels.transpose()[c] > (2*thres_vector[c]) for c in range(19)]).transpose()
+    sc_labels = il_labels*sc_labels 
+    print(sc_labels.shape)
     sc_labels = pd.DataFrame(sc_labels.astype('uint8'))
     sc_labels.columns = LABEL_ALIASE_LIST
     # Merge the calculated sc_labels back with meta
@@ -305,7 +306,7 @@ def prepare_meta_publicHPAv21():
     df_c.target.value_counts()
     print('Keeping these final columns:')
     print(df_c.columns)
-    df_c.to_csv(f'{d}/sl_pHPA_15_0.05_euclidean_100000_rmoutliers_ilsc_3d_bbox_metav21_individualthresholds.csv', index=False)
+    df_c.to_csv(f'{d}/sl_pHPA_15_0.05_euclidean_100000_rmoutliers_ilsc_3d_bbox_metav21_individualthresholds_il.csv', index=False)
     #df_c.to_csv(f'{d}/sl_pHPA_15_0.05_euclidean_100000_rmoutliers_ilsc_3d_bbox_metav21_potentialnewlabels.csv', index=False)