search

AdneneBoumessouer / MVTec-Anomaly-Detection

This project proposes an end-to-end framework for semi-supervised Anomaly Detection and Segmentation in images based on Deep Learning.
278 stars 59 forks source link

The detail information of reconstruction map is fuzzy #15

Closed mozhuqing closed 4 years ago

mozhuqing commented 4 years ago

Hello, I have a question. Why is the reconstructed image missing the details? For example, hazelnut, whose reconstruction details are fuzzy. The details of the pictures you have made are also some true. The words "500" on the capsule are very vague.

AdneneBoumessouer commented 4 years ago

One limitation of the Auto-Encoder is the unability to reconstruct some details that are present only in a very small subset of the total training image dataset, like the 500 observed in some pill-images. I have considerably improved the code. Try working with the new version.