![[Neurocle] Logo_white.png](https://static.wixstatic.com/media/242158_0a4304ae861e4107925bc861d88c280f~mv2.png/v1/fill/w_150,h_21,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/%5BNeurocle%5D%20Logo_white.png){kind=small}
New Features
New Software 4.0 Version Release!
GAN Model &
Generation Center
Generate the realistically similar virtual defects and use them to train inspection models. When the model is trained using these generated virtual defect images, a significantly improved inspection accuracy can be achieved compared to the existing model.
The Generation Center provides various options for generating defects, allowing users to create a diverse range of defects that closely resemble real-world defects, according to their preferences.
.png)
Unsupervised Model :
Anomaly Classification & Anomaly Segmentation
Make inspection model that can define OK / NG by training only with normal images by minimal labeling resources.
Neurocle's unsupervised learning models boast high accuracy levels.
Furthermore, the these models provide a visual indicator of the basis for categorizing images as abnormal. This is accomplished through heatmaps and blobs, offering users a more detailed outcome.
Smart Labeling Tools :
Auto-Selector & Keyword Labeler
Auto-Selector is a feature that automatically identifies areas within an image based on the objects present, targeting regions with similar attributes among objects.
Keyword Labeler labels portions of an image with bounding boxes that corresponds upon input keywords. To label more elaborately, the 'Convert Boxes to Polygons' function allows the transformation of bounding box labels into polygonal shapes for greater precision.
.png)
Multi-model Export
Multiple models are interconnected in parallel within a Flowchart model, which can be invoked and utilized through a single API during runtime.
This approach substantially reduces the programming effort required during runtime, leading to significant time savings.
.png)
High-level API Module
The Real-time Inference Engine, Neuro-R's API, has been enhanced.
Users can leverage the advanced API to drastically reduce the programming effort involved in the process of model application.
Pretrained OCR
Create highly accurate models without the need for any manual labeling effort.
In comparison to models generated after manual labeling, this approach maintains inspection accuracy while drastically reducing the time required for model creation.
