Automated Report Generation with Computer Vision

Team track record (prior role) · AI / Team track record

An end-to-end pipeline that automates report generation from images and documents.

From our team's track record: developed by our AI lead in a previous data science role. This was not a NezT Digital client engagement.

What our ai lead owned

The report-generation pipeline and its three micro-pipelines.

Evaluating 36 image-classification models and selecting the best performers.

Training an oriented object detection model and combining OCR with an LLM for structured output.

Tech stack

Python, ResNet50V2, DenseNet169, YOLOv8, OCR, LangChain, Roboflow

Impact

96.8% test accuracy on image classification and mAP@.50 of 0.98 on object detection, replacing a manual extraction and assembly workflow.