From e328910fd29bb1851e041db7a050686245a62e62 Mon Sep 17 00:00:00 2001
From: aufr33 <65520685+aufr33@users.noreply.github.com>
Date: Wed, 7 Jul 2021 09:26:57 +0300
Subject: [PATCH] Update dataset.py

---
 lib/dataset.py | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/lib/dataset.py b/lib/dataset.py
index 5a049b3..bdeebc7 100644
--- a/lib/dataset.py
+++ b/lib/dataset.py
@@ -79,24 +79,26 @@ def train_val_split(dataset_dir, split_mode, val_rate, val_filelist):
     return train_filelist, val_filelist
 
 
-def augment(X, y, reduction_rate, reduction_mask, mixup_rate, mixup_alpha):
+def augment(X, y, reduction_rate, reduction_mask, mixup_rate, mixup_alpha, mp):
     perm = np.random.permutation(len(X))
     for i, idx in enumerate(tqdm(perm)):
         if np.random.uniform() < reduction_rate:
             y[idx] = spec_utils.reduce_vocal_aggressively(X[idx], y[idx], reduction_mask)
-
-        if mixup_rate > 0.0:
+        
+        if not any([mp.param["mid_side"], mp.param["mid_side_b"], mp.param["mid_side_c"]]):
             if np.random.uniform() < 0.5:
                 # swap channel
                 X[idx] = X[idx, ::-1]
                 y[idx] = y[idx, ::-1]
+                    
             if np.random.uniform() < 0.02:
                 # mono
                 X[idx] = X[idx].mean(axis=0, keepdims=True)
                 y[idx] = y[idx].mean(axis=0, keepdims=True)
-            if np.random.uniform() < 0.02:
-                # inst
-                X[idx] = y[idx]
+                
+        if np.random.uniform() < 0.02:
+            # inst
+            X[idx] = y[idx]
 
         if np.random.uniform() < mixup_rate and i < len(perm) - 1:
             lam = np.random.beta(mixup_alpha, mixup_alpha)