7 PM – AI/ML – Darin

Today’s Activities:

1. Finalized finals projects, debugged streamlit and dataset related issues.
2. Finished the Boston Housing Price Prediction!

Homework:

1. Get your app working on the streamlit website! One more thing you have to ensure is that the labeling function exists with the same name on your pycharm script
2. Watch this 45 mins recording about Random Forest https://www.simplilearn.com/tutorials/machine-learning-tutorial/random-forest-algorithm

For Reine, this code is the fix, I have tested and verified training:

path = "/kaggle/input/datasets/vesuvius13/formula-one-cars/Formula One Cars"
# Get every image under all team folders
all_files = get_image_files(path)
print("Total files before cleaning:", len(all_files))

from PIL import Image
from pathlib import Path

def is_valid_image(fn):
    try:
        # First check the image header
        with Image.open(fn) as im:
            im.verify()

        # Then actually load/convert it, because verify() alone can miss some bad files
        with Image.open(fn) as im:
            im.convert("RGB").load()

        return True

    except Exception as e:
        return False

good_files = []
bad_files = []

for fn in all_files:
    if is_valid_image(fn):
        good_files.append(fn)
    else:
        bad_files.append(fn)

print("Good files:", len(good_files))
print("Bad files:", len(bad_files))

for fn in bad_files[:30]:
    print(fn)

def extract_brand(fn):
    folder_name = Path(fn).parent.name
    return folder_name.replace(" F1 car", "").strip()

print(extract_brand("/kaggle/input/datasets/vesuvius13/formula-one-cars/Formula One Cars/Racing Point F1 car/00000090.png"))

Notes:

You can reach me at ddjapri@ayclogic.com.

All class notes can be found here.