data aug: random background, write summary of CERs while training

RajChauhan08 · Feb 2, 2021 · cdfb518 · cdfb518
1 parent 2f0499f
commit cdfb518
Show file tree

Hide file tree

Showing 9 changed files with 25 additions and 17 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1,8 +1,6 @@
 data/words*
 data/words.txt
 src/__pycache__/
-model/checkpoint
-model/snapshot-*
 notes/
 *.so
 *.pyc

diff --git a/README.md b/README.md
@@ -21,7 +21,6 @@ The input image and the expected output is shown below.
 
 ```
 > python main.py
-Validation character error rate of saved model: 11.118344571029994%
 Init with stored values from ../model/snapshot-76
 Recognized: "Hello"
 Probability: 0.8462573289871216

diff --git a/data/pixelRelevance.npy b/data/pixelRelevance.npy
diff --git a/data/translationInvariance.npy b/data/translationInvariance.npy
diff --git a/data/translationInvarianceTexts.pickle b/data/translationInvarianceTexts.pickle
diff --git a/model/.gitignore b/model/.gitignore
@@ -0,0 +1,4 @@
+# Ignore everything in this directory
+*
+# Except this file
+!.gitignore
diff --git a/model/accuracy.txt b/model/accuracy.txt
diff --git a/src/SamplePreprocessor.py b/src/SamplePreprocessor.py
@@ -23,30 +23,30 @@ def preprocess(img, imgSize, dataAugmentation=False):
         if random.random() < 0.25:
             img = cv2.erode(img, np.ones((3, 3)))
         if random.random() < 0.5:
-            img = img * (0.5 + random.random() * 0.5)
+            img = img * (0.1 + random.random() * 0.9)
         if random.random() < 0.25:
-            img = np.clip(img + (np.random.random(img.shape) - 0.5) * random.randint(1, 50), 0, 255)
+            img = np.clip(img + (np.random.random(img.shape) - 0.5) * random.randint(1, 25), 0, 255)
         if random.random() < 0.1:
             img = 255 - img
 
         # geometric data augmentation
         wt, ht = imgSize
         h, w = img.shape
         f = min(wt / w, ht / h)
-        fx = f * np.random.uniform(0.75, 1.25)
-        fy = f * np.random.uniform(0.75, 1.25)
+        fx = f * np.random.uniform(0.75, 1.5)
+        fy = f * np.random.uniform(0.75, 1.5)
 
         # random position around center
         txc = (wt - w * fx) / 2
         tyc = (ht - h * fy) / 2
-        freedom_x = wt // 10
-        freedom_y = ht // 10
-        tx = txc + np.random.randint(-freedom_x, freedom_x)
-        ty = tyc + np.random.randint(-freedom_y, freedom_y)
+        freedom_x = wt / 5
+        freedom_y = ht / 5
+        tx = txc + np.random.uniform(-freedom_x, freedom_x)
+        ty = tyc + np.random.uniform(-freedom_y, freedom_y)
 
         # map image into target image
         M = np.float32([[fx, 0, tx], [0, fy, ty]])
-        target = np.ones(imgSize[::-1]) * 255 / 2
+        target = np.ones(imgSize[::-1]) * np.random.uniform(0, 255)
         img = cv2.warpAffine(img, M, dsize=imgSize, dst=target, borderMode=cv2.BORDER_TRANSPARENT)
 
     # no data augmentation

diff --git a/src/main.py b/src/main.py
@@ -1,4 +1,5 @@
 import argparse
+import json
 
 import cv2
 import editdistance
@@ -12,14 +13,20 @@
 class FilePaths:
     "filenames and paths to data"
     fnCharList = '../model/charList.txt'
-    fnAccuracy = '../model/accuracy.txt'
+    fnSummary = '../model/summary.json'
     fnInfer = '../data/test.png'
     fnCorpus = '../data/corpus.txt'
 
 
+def write_summary(charErrorRates):
+    with open(FilePaths.fnSummary, 'w') as f:
+        json.dump(charErrorRates, f)
+
+
 def train(model, loader):
     "train NN"
     epoch = 0  # number of training epochs since start
+    summaryCharErrorRates = []
     bestCharErrorRate = float('inf')  # best valdiation character error rate
     noImprovementSince = 0  # number of epochs no improvement of character error rate occured
     earlyStopping = 25  # stop training after this number of epochs without improvement
@@ -39,14 +46,16 @@ def train(model, loader):
         # validate
         charErrorRate = validate(model, loader)
 
+        # write summary
+        summaryCharErrorRates.append(charErrorRate)
+        write_summary(summaryCharErrorRates)
+
         # if best validation accuracy so far, save model parameters
         if charErrorRate < bestCharErrorRate:
             print('Character error rate improved, save model')
             bestCharErrorRate = charErrorRate
             noImprovementSince = 0
             model.save()
-            open(FilePaths.fnAccuracy, 'w').write(
-                f'Validation character error rate of saved model: {charErrorRate * 100.0}%')
         else:
             print(f'Character error rate not improved, best so far: {charErrorRate * 100.0}%')
             noImprovementSince += 1
@@ -140,7 +149,6 @@ def main():
 
     # infer text on test image
     else:
-        print(open(FilePaths.fnAccuracy).read())
         model = Model(open(FilePaths.fnCharList).read(), decoderType, mustRestore=True, dump=args.dump)
         infer(model, FilePaths.fnInfer)