renaming all files - moving training to be a single file for transfer vs not transfer learning. Made the testing file test all models. Needs to be updated to only update with new models.

show_incorrect_images.py

@@ -0,0 +1,35 @@
+import pandas as pd
+import matplotlib.pyplot as plt
+import matplotlib.image as mpimg
+
+df = pd.read_csv("sub1_non_transfer.csv")
+df2 = pd.read_csv("poke_evos.csv")
+
+evos = []
+
+for index, row in df2.iterrows():
+    print(row)
+    s = ""
+    s+=row["stage1"] if not pd.isnull(row["stage1"]) else ""
+    s+=row["stage2"] if not pd.isnull(row["stage2"]) else ""
+    s+=row["stage3"] if not pd.isnull(row["stage3"]) else ""
+    evos.append(s.lower().replace(" ", "-").rstrip())
+
+
+incorrect = df[df["prediction"]!= df["true_val"]]
+
+total_same_fam = 0
+# TODO: Add in support for figuring out if the pokemon are related/evolutions of one another
+for index, row in incorrect.iterrows():
+    img = mpimg.imread("./SingleImageTestSet/" + row['fname'])
+    imgplot = plt.imshow(img)
+    title = f"Predicted - {row['prediction']}, Actual - {row['true_val']}"
+    for evo in evos:
+        if row['prediction'] in evo and row['true_val'] in evo:
+            title+=f"\n same family name detected - {evo}"
+            total_same_fam+=1
+    plt.title(title)
+    plt.show()
+
+
+print(f"The total number of incorrect entries from same families is {total_same_fam} - {total_same_fam/len(incorrect)}")