TrashGuard Dashboard
Project Idea

TrashGuard

Spot the dump, snap the pic, help keep San José clean.

Selected track: Clean up your city

💡 Inspiration

As an international student new to San José, I experienced the city with fresh eyes. Near campus I kept seeing overflowing bins and illegally dumped mattresses and furniture.
I soon learned about San José’s 311 system, a channel for non-emergency service requests—a civic tool I’d never used back home.

Yet public data show illegal-dumping reports are rising, while anecdotes suggest cleanup crews still follow fixed routes that may miss brand-new piles.
So I asked:

Can we augment 311 with a focused app that turns resident snapshots into a live hotspot map—helping the City direct crews faster?

🌍 What TrashGuard Does

Feature	Details
Maps Current Intensity	Google-Maps heat-map of every “Illegal Dumping” 311 ticket from the last 90 days (via Open-Data CKAN).
Highlights Persistent Areas	Aggregates tickets into 200 m grid tiles—recurring hotspots glow brighter.
Focused Citizen Reporting	Users: • Upload a photo • Choose size (Small/Med/Large) & type (Furniture, Bags…) • Capture GPS with one tap.
Persistent Cloud Storage	• Metadata → Supabase (Postgres) • Images → Google Cloud Storage
Admin Workflow	Password-protected sidebar lets staff view reports, change status (“New / Reviewed / Cleaned”), or delete errors.

❓ Why another uploader when 311 already accepts photos?

Extra value from TrashGuard	Why 311 alone isn’t always enough	Status
Visual context on a live map	311 form shows no heat-map; TrashGuard gives instant context & confirmation.	✅
GPS precision	Mistyped 311 addresses default to city centroid; we use device GPS.	✅
Structured severity data	311 text is generic; we collect Size/Type plus photo.	✅
Single-purpose UX	311 covers dozens of services; TrashGuard is dumping-only, super-fast.	✅
Enriched analytics fields	Adds structured size/type/location/status & direct image link.	✅
Foundation for duplicate detection	Precise GPS + timestamp + photo enable auto-merge later.	Enabled
Gamified lifecycle	Watch your marker change status live—future enhancement.	Future

Think of TrashGuard as a specialised overlay that enriches 311 data and visualises urgency.

🛠️ How I Built It

Layer	Tech	Key Steps
Data Retrieval	Python + CKAN SQL API	Query “Service Type = 'Illegal Dumping'”.
Processing	Pandas / NumPy	Parse “Date Created”, filter 90 days, clean lat/lon, bin to 200 m tiles.
Caching	`@st.cache_data`	Speed up reloads, cut API calls.
Map UI	Streamlit + HTML component	Embed Google Maps HeatmapLayer; tune radius & maxIntensity.
Geo-Location	`streamlit-js-eval`	Grab `navigator.geolocation` coords.
Metadata DB	Supabase (Postgres)	CRUD for reports + status updates.
Photo Storage	Google Cloud Storage	Upload via service account, retrieve public URL.
Admin Panel	Streamlit sidebar	List reports, show image, status dropdown, delete button.
Modularity	Python packages	`data_handler`, `db_handler`, `storage_handler`, `utils`.
Deploy	Streamlit Cloud	Secrets for Maps, CKAN, Supabase, GCS in `.streamlit/secrets.toml`.

📚 What I Learned

Cloud integration: Streamlit can talk to Supabase + GCS with service keys.
GCS IAM nuances: Uniform vs fine-grained ACLs, public read vs backend write.
Supabase Python client: Great for CRUD but needs careful error handling.
streamlit-js-eval bridges browser APIs (GPS) back to Python.
Modular design keeps the codebase sane and testable.
st.session_state is vital for sticky messages across st.rerun().