flat-finder

As of late 2022, serious rewrite from BeautifulSoup to Selenium is needed because BeautifulSoup is struggling to download the whole page. This results in just 5-10% images downloaded.

---

Looking for a modern, nicely looking apartment but tired with all these old flats from the '90s? Let's automate that :)

Scrape apartment offers from OLX․pl, analyse them using artificial intelligence (AI) model to get only good looking flats and (optionally) run IFTTT automation (eg. send email; add a to-do task) when new offers matching search criteria (eg. ≥ 2 rooms, ≤ 2k PLN, ≥ 30m²) are found.

With support for native macOS & Windows 10 notifications.

Screenshots

Features

• Get offers from OLX․pl
• Pagination support
• Artificial intelligence (AI) model to assess if an aparment is modern or not
• IFTTT (eg. send an email or add a to-do task if new offer is found)
• Cross platform (macOS & Windows)
• Native notifications (macOS & Windows)

How to use

Coming soon™.

Release History

• 0.14.2: Fixed offer counter & Windows 10 notification icon not showing up.
• 0.14.1: Added two additional calculations to know how much time it takes to train the model and how much time it takes to download and classify the images.
• 0.14: Refactored code; moved some more code to functions; added support for more than one URL with offers (eg. useful w/ multiple locations); tried to achieve feature parity between Windows and macOS; improved model accuracy; changed comparison mode so that previously seen offers are not shown again; cleaner output / general look & feel; removed unused code; other small improvements.
• 0.13: Open folders for manual review of the images.
• 0.12: Added colorama to have colored output; changed URL; added more images to model's dataset; added fix for feedModel.py crashing due to duplicates found in target location.
• 0.11.3: Forced to look only on 3 subpages.
• 0.11.2: Increased # of epochs to improve model's accuracy; added more images to model's dataset.
• 0.11.1: Changed title in notifications.
• 0.11: Automated runs using Task Scheduler (Windows) & Automator (macOS). Added more images to model's dataset. Disabled IFTTT automation for testing.
• 0.10.2: Cleaned up the code. Added comments.
• 0.10.1: Cleaned up the code.
• 0.10: Added macOS compatibility. Enabled IFTTT automation. Improved model accuracy - newly downloaded images are used to feed the model so it can improve itself over time.
• 0.9: Pagination support.
• 0.8: Enabled Windows 10 & macOS notifications when new offers are found.
• 0.7: Downloading images to unique folders. Added data comparison logic to show which offers are new vs previous script run.
• 0.6: Adding only "modern" offers to a text file. Changed local images renaming logic.
• 0.5: Started to build some if/else logic for when offer is modern/ancient.
• 0.4: Added v1 of AI model to assess which offer shows new ("modern") and which old ("ancient") apartments.
• 0.3: Downloading offers' main images'.
• 0.2: Getting offers' URLs + offers' main images' URLs.
• 0.1: Initial release.

Versioning

Using SemVer.

License

Acknowledgements

Libs

• TensorFlow
• BeautifulSoup
• alive-progress
• win10toast
• win10toast-persist
• win10toast-click
• pync
• GD Shortener
• wget
• Pillow
• NumPy
• Matplotlib
• colorama

Stack Overflow

• certificate issue fix
• click Windows 10 notification to open URL

Other

• Images used to train the model were downloaded from Google Images. Respective licenses apply.
• Flaticon / Freepik
• IFTTT
• Connect a Python Script to IFTTT by Enrico Bergamini
• Use IFTTT web requests to send email alerts by Anthony Hartup
• Image classification

Contributing

If you found a bug or want to propose a feature, feel free to visit the Issues page.