./construction_progress

Intelligent Construction Progress Tracking Pipeline

Internship project · Computer Vision + Multimodal AI

The problem

Construction sites generate huge volumes of field imagery, but turning those photos into a clear “what happened today” narrative was manual and slow.

Approach

Built a pipeline that:

  1. Classifies site imagery with object detection (YOLO)
  2. Interprets each image with a vision-language model (Qwen2.5-VL)
  3. Generates a structured progress narrative via an LLM

Stack

  • YOLO
  • Qwen2.5-VL
  • LLM pipeline

Outcome

Enabled structured, faster construction progress reporting directly from field images — cutting down the manual effort of translating photos into reports.

[ screenshots + repo link coming soon ]