AutoInspect AI

Inspiration

In manufacturing, infrastructure maintenance, and retail, subtle visual changes — like a missing bolt, surface crack, or faded logo — can have major consequences. Manual inspection is slow, error-prone, and inconsistent. I wanted to create a system that automates visual inspections using AI, saving time and improving accuracy.

What it does

AutoInspect AI analyzes time-series images to detect and classify changes. It identifies structural damage, surface wear, or brand compliance issues and generates actionable reports with heatmaps and severity scores.

How we built it

Languages: Python, JavaScript (for dashboard)

Frameworks/Libraries: PyTorch/TensorFlow, OpenCV, NumPy/Pandas, Streamlit/React

Cloud Services: AWS S3 / GCP Storage for images, AWS SageMaker / GCP AI Platform for model deployment

Databases: PostgreSQL or MongoDB for storing images, metadata, and reports

APIs: REST API for image ingestion and notifications

The system uses AI embeddings rather than raw pixel comparison, making it robust to lighting, angle, and scale variations.

Challenges we ran into

Handling differences in lighting, angle, and scale across images

Designing semantic-aware change detection instead of simple pixel comparison

Prioritizing changes by severity for actionable insights

Accomplishments that we're proud of

Conceptualized a general-purpose visual difference engine applicable across industries

Designed a system that is robust, scalable, and explainable

Proposed actionable outputs like heatmaps and severity scoring for inspections

What we learned

Importance of semantic embeddings in computer vision for detecting meaningful changes

How to align and preprocess time-series images for consistent analysis

Techniques for building scalable AI pipelines with real-world applications

What's next for AutoInspect AI

Integrate real-time drone or CCTV feeds for continuous monitoring

Expand to multimodal inputs like sensor data alongside images

Add predictive maintenance insights based on detected changes over time

Built With

• and-reports)-apis:-optional-rest-api-for-image-ingestion
• aws-s3/gcp-storage
• aws-sagemaker-or-gcp-ai-platform-(model-deployment)-databases:-postgresql-or-mongodb-(storing-images
• aws-sagemaker-or-gcp-ai-platform-databases:-postgresql-or-mongodb-apis:-rest-api-for-image-ingestion-and-notifications-other-tools:-onnx-or-tensorrt-for-edge-deployment
• aws-sagemaker/gcp-ai-platform
• javascript
• javascript-(for-dashboard)-frameworks/libraries:-pytorch-or-tensorflow-(for-cv-models)
• javascript-frameworks/libraries:-pytorch/tensorflow
• metadata
• notifications
• numpy
• numpy/pandas-(data-handling)
• onnx/tensorrt
• opencv
• opencv-(image-processing)
• pandas
• postgresql/mongodb
• python
• pytorch
• pytorch/tensorflow
• react
• rest-api
• streamlit
• streamlit-or-react-(dashboard/prototype-ui)-cloud-services:-aws-s3-or-google-cloud-storage-(image-storage)
• streamlit/react
• streamlit/react-cloud-services:-aws-s3-or-google-cloud-storage
• tensorflow