AI for Occupancy Analytics | Building a Smart Parking System

In this video, João Marcos, Product Engineer at Roboflow, demonstrates how to build a computer vision application to manage logistics in an intermodal yard. You will learn how to process drone footage to identify which parking stalls are empty and which are occupied by trailers.

João walks through the entire process, starting with Roboflow Rapid to detect trailers and stall numbers without labeling any data. He then dives into Roboflow Workflows to architect a logic solution for determining occupancy. Finally, he integrates an OCR model (OpenAI) to read the specific stall numbers for actionable reporting.

= Topics covered in this video =
⚡ Rapid Prototyping: How use text prompts to detect objects without labeling and training a model from scratch.
↔️ Workflow Logic: Using "horizontal padding" to verify if a trailer is parked next to a stall number.
⚖️ Advanced Filtering: Using Detection Consensus to identify empty stalls by comparing datasets.
🔡 OCR Integration: cropping, rotating, and sending images to an OCR model to read text.

= Chapters =
00:00 Introduction: Visualizing Occupied vs. Empty Stalls
00:43 Architecting the Solution & Logic
02:42 Detecting Trailers & Numbers with Roboflow Rapid
04:15 Setting up the Roboflow Workflow
08:29 Using Horizontal Padding to Understand Occupancy
10:00 Filtering for Occupied Stalls (Overlap Filter)
12:51 - Identifying Empty Stalls (Detection Consensus)
15:58 - Reading Stall Numbers with OCR & Pre-Processing
19:30 - Final Results & Batch Processing Visualization

= Resources =
Get started with Roboflow Rapid: https://rapid.roboflow.com/

Fork the Workflow you saw in this video:
https://app.roboflow.com/workflows/em...