analysis on all possible files in a provided directory, writes to JSON

.gitignore

@@ -0,0 +1,2 @@
+__pycache__
+predictions

main.py

@@ -3,23 +3,46 @@
 # most of this application adapted from the following walkthrough:
 # https://towardsdatascience.com/how-to-use-a-pre-trained-model-vgg-for-image-classification-8dd7c4a4a517
 
-import sys, os
+import sys, os, json, time
 from predict import predict
 from keras.applications.vgg16 import VGG16
 
+print("\n\n\n")
+print("Imports successful! Running startup processes...")
+
+# generate current time for use in identifying outfiles
+cur_time = str(int(time.time()))
+
+# create the target directory if it doesn't exist
+if (not os.path.exists("./predictions")):
+    print("Did not find predictions directory, creating...")
+    os.makedirs("./predictions")
+
 # declare model to be used for each prediction
 model = VGG16(weights='imagenet')
 
 # receive directory path as CLI argument and get a list of all files in path
 path = sys.argv[1]
+if (path[-1] != "/"):
+    path += "/"
+
 files = os.listdir(path)
 
 # store all results in one list
 all_results = []
 
+print("Running image analysis. This may take some time")
+
 # for each file in directory, append its prediction result to main list
 for file in files:
-    result = predict(model, file)
+    result = predict(model, path + file)
-    all_results.append({ path: file, result: result })
+    if result is not None:
+        all_results.append({ "path": file, "prediction": result })
 
-print(all_results)
+print("Analysis complete! Writing JSON to ./predictions/predictions" + cur_time + ".json")
+
+# convert object to JSON and write to JSON file
+with open("./predictions/predictions" + cur_time + ".json", "w") as outfile:
+    json.dump(all_results, outfile)
+
+print("Process complete!")

predict.py

@@ -3,8 +3,12 @@ from keras.utils import load_img, img_to_array
 from keras.applications.vgg16 import preprocess_input, decode_predictions
 
 def predict(model, path):
+    # only allow valid file types
+    if not (".jpg" in path or ".jpeg" in path):
+        return None
+
     # receive image path as CLI argument
-    img = load_img(path ,color_mode='rgb', target_size=(224, 224))
+    img = load_img(path, color_mode='rgb', target_size=(224, 224))
 
     # loaded image to np array for model to read
     x = img_to_array(img)
@@ -16,4 +20,10 @@ def predict(model, path):
     features = model.predict(x)
     p = decode_predictions(features)
     
+    for predict in p:
+        i = 0
+        while i < len(predict):
+            predict[i] = (predict[i][0], predict[i][1], str(predict[i][2]))
+            i = i + 1
+
     return p

readresult.py

@@ -0,0 +1,24 @@
+import sys, json
+
+path = sys.argv[1]
+
+with open(path) as file:
+    contents = json.load(file)
+
+for line in contents:
+    prediction = line['prediction']
+    for section in prediction:
+        for guess in section:
+            if (float(guess[2]) > 0.75):
+                print(line['path'])
+                print("Probable match: " + guess[1])
+                print(guess)
+                print("\n")
+            elif (float(guess[2]) > 0.3):
+                print(line['path'])
+                print("Potential match: " + guess[1])
+                print(guess)
+                print("\n")
+            # else:
+            #     print(line['path'] + ": inconclusive")
+            #     print("\n")