From 4de74afcd15993574294b5011789b03b198cfc7c Mon Sep 17 00:00:00 2001
From: raphael <raphael@maenle.net>
Date: Fri, 6 Aug 2021 13:24:08 +0200
Subject: [PATCH] adds coco image split script, update evaluation

- evaluation now pauses via pdb and correctly divides
  the image sizes
- coco image requires a 'coco' folder with the COCO
  'val2014' and 'annotations' folder downloaded from the
  cocodataset website
- the script splits up the dataset by snipping the bounding-
  box from the images and saving it into a seperate folder
  for later parsing by the siamese network
---
 coco_gen.py | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 evaluate.py | 20 ++++++++++-----
 2 files changed, 88 insertions(+), 6 deletions(-)
 create mode 100644 coco_gen.py

diff --git a/coco_gen.py b/coco_gen.py
new file mode 100644
index 0000000..bb473e1
--- /dev/null
+++ b/coco_gen.py
@@ -0,0 +1,74 @@
+import json, os
+import matplotlib.pyplot as plt
+import numpy as np
+
+from PIL import Image
+
+
+no_label = 0
+small = 0
+passed = 0
+
+count = 0
+
+
+print("loading coco annotations...")
+coco = json.load(open('./coco/annotations/instances_val2014.json'))
+print("done")
+
+def findAnnotationName(annotationId):
+    for c in coco['categories']:
+        if c['id'] == annotationId:
+            return c['name']
+
+def findAnnotationToId(ident):
+    for annotation in coco['annotations']:
+        img_an = annotation['image_id']
+        if img_an == ident:
+            return annotation 
+
+def show(pil, pause=0.2):
+    ImageNumpyFormat = np.asarray(pil)
+    plt.imshow(ImageNumpyFormat)
+    plt.draw()
+    plt.pause(pause) # pause how many seconds
+    plt.close()
+
+
+def parseImage(coImg):
+    # open image file
+
+    path = "coco/val2014/" + coImg['file_name']
+    img = Image.open(path)
+
+    an = findAnnotationToId(coImg['id'])
+    if an == None:
+        no_label += 1
+        return
+
+
+    path = "coco/val2014/" + coImg['file_name']
+    img = Image.open(path)
+    c = an['bbox']
+    crop = img.crop((c[0], c[1], c[0]+c[2], c[1]+c[3]))
+
+    if crop.width < 64 or crop.height < 64:
+        small += 1
+        return
+
+    imagePath = f"classified/{findAnnotationName(an['category_id'])}/{an['id']}.png"
+    os.makedirs(os.path.dirname(imagePath), exist_ok=True)
+    crop.save(imagePath)
+    passed += 1
+
+if __name__ == "__main__":
+    for coImg in coco['images']:
+        parseImage(coImg)
+        count += 1
+        if count % 100 == 0:
+            print("status:")
+            print(f"no labels: {no_label}")
+            print(f"to small: {small")
+            print(f"passed: {passed}")
+            print("-----")
+
diff --git a/evaluate.py b/evaluate.py
index ae55361..ec452cb 100644
--- a/evaluate.py
+++ b/evaluate.py
@@ -2,12 +2,20 @@ import tensorflow.keras as keras
 from PIL import Image
 import numpy as np
 import pdb
-model = keras.models.load_model('./siamese_checkpoint')
-image1 = np.asarray(Image.open('../towards/data/fruits-360/Training/Avocado/r_254_100.jpg').convert('RGB').resize((100,
-    100))) / 255
-image2 = np.asarray(Image.open('../towards/data/fruits-360/Training/Avocado/r_250_100.jpg').convert('RGB').resize((100,
-    100))) / 255
 
-output = model.predict([np.array([image2]), np.array([image1])]) 
+def getI(path):
+    return np.asarray(Image.open(path).convert('RGB').resize((100, 100))) / 255
+
+def predict(image1, image2):
+    return model.predict([np.array([image2]), np.array([image1])]) 
+
+model = keras.models.load_model('./siamese_checkpoint')
+image1 = getI('../towards/data/fruits-360/Training/Avocado/r_254_100.jpg')
+image2 = getI('../towards/data/fruits-360/Training/Avocado/r_250_100.jpg')
+
+print(predict(image1, image2))
+
 pdb.set_trace()
 
+
+