Computer Vision

Image models outside the benchmark: robustness, adversarial attacks, synthetic training data, and evaluation on real images.

October 1, 2026 Computer Vision

Can a Segmentation Model Catch Mistakes in Its Own Training Labels?

A nuclei segmentation pipeline on public microscopy data, from checking the hand-drawn masks to a web app that runs on a CPU, and what the trained model found in the labels it learned from.

Computer Vision Image Segmentation Data Labeling Transfer Learning Model Evaluation ONNX PyTorch Deep Learning Microscopy

June 6, 2026 Computer Vision

Can Fake Training Data Make AI Harder to Fool? A Traffic Sign Experiment

A traffic sign experiment on whether synthetic training data can improve adversarial robustness without sacrificing accuracy on real images

Computer Vision Synthetic Data Robustness AI safety Generative AI