March 19, 2026 | News | Deep Learning Tool
The Deep Learning Tool 26.03 introduces a major upgrade for Deep OCR. The new Deep OCR Alignment feature allows you to perform both the detection and recognition steps in a single pass. This works for single-line text or roughly cropped images, eliminating the need for a separate detection step, which significantly boosts inference speed.
Additionally, several other improvements enhance Deep OCR performance. New recognition and detection models enable faster inference times with reduced memory requirements. The image width is now auto-detected during training, making the process more efficient. Training has also been optimized, resulting in shorter training times and lower memory usage.
Another key benefit is the ability to evaluate MVTec’s pretrained models directly in the Deep Learning Tool without the need for any training. Moreover, the latest GPU models with CUDA 12.8 support are now fully compatible.
These updates make MVTec's Deep OCR faster, more efficient, and easier to deploy.
You can download and install the new version using the MVTec Software Manager (SOM). Be sure to click “refresh from server” first to get the most up-to-date list of available versions. For users who cannot use the online installer, we also provide an offline installer on the MVTec website.
Published on: March 19, 2026
