Contact person: Business Client

Phone:Show

Email:Show

Location: FAISALABAD, Pakistan

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I need an AI-driven application that reviews photos of residential buildings and highlights two specific problem classes: structural issues (cracks, subsidence, spalling, etc.) and water damage (damp patches, mold traces, leakage marks). Image analysis is the sole detection method, so the heart of the job is a well-trained computer-vision model—CNNs, Vision Transformers, or any modern approach you prefer—as long as it reliably tags each defect and shows its location on the photo.

Here is the workflow I am aiming for:
1. A homeowner, inspector, or contractor snaps or uploads exterior or interior images.
2. The system runs inference, draws bounding boxes or masks around suspected defects, and labels them with confidence scores.
3. A short PDF or JSON report summarises the findings per image.

For training, I can supply a starter set of annotated pictures; you are free to augment the data or tap into open-source datasets and transfer learning to reach robust performance. The deliverables I expect are:

• Reproducible training code in Python (TensorFlow or PyTorch).
• A Dockerised inference API that works on my Windows laptop and an AWS EC2 instance.
• A lightweight web or mobile demo front-end that calls the API and displays results.
• Precision and recall above 90 % on a validation set I will provide.

Hit those marks and we have a successful MVP. If things go smoothly, future phases may expand to other defect types, so write clean, modular code ready for new classes later on. I’m ready to start as soon as you are." (client-provided description)