About Us
GrayMor Tech
C++ and only C++
C++ and only C++
GrayMor Tech is founded by two computer geeks, one has 35years of C/C++ the other 12 years and a Phd in machine learning. We built this because we though it needed to exist.
If, like us, you needed to add DNN image segmentation/classification to a device that can't go off to the internet to process or run some Python script or call into Ma
GrayMor Tech is founded by two computer geeks, one has 35years of C/C++ the other 12 years and a Phd in machine learning. We built this because we though it needed to exist.
If, like us, you needed to add DNN image segmentation/classification to a device that can't go off to the internet to process or run some Python script or call into MatLab, then we may have the answer :-)
it's compact and just works!
A DLL or .SO and a header file !
C++ and only C++
C++ and only C++
C++ and only C++
This App/library is for C++ developers. It's built on Cuda, cuDNN, MKL &TBB.
There is no Python, TensorFlow, PyTorch or anything else to learn.
Delve++ was built so regular C++ devs could add DNNs for image processing without becoming a machine learning guru first.
Smart GPU memory management means it will put as much of your net on the GPU
This App/library is for C++ developers. It's built on Cuda, cuDNN, MKL &TBB.
There is no Python, TensorFlow, PyTorch or anything else to learn.
Delve++ was built so regular C++ devs could add DNNs for image processing without becoming a machine learning guru first.
Smart GPU memory management means it will put as much of your net on the GPU as it can, but use the CPU if it has to.
Design build and train on windows 10/11, then deploy to your own apps by linking to the DeepCEval library
The trained network archive can be used on both Windows or Linux. The evaluation lib uses as little GPU memory as possible, allowing multiple nets to be loaded and run at once.
WHY Delve++ ?
C++ and only C++
Quick Start Res-Net
We have some handy layer types, like a RANSAC loss layer that directly outputs circular splines. Very useful if your trying to find a particular surface in an image.
We develop layers ourselves based on the latest research papers.
Because we only support C++ there's no overhead or compromise to accommodate other languages. This one is by C+
We have some handy layer types, like a RANSAC loss layer that directly outputs circular splines. Very useful if your trying to find a particular surface in an image.
We develop layers ourselves based on the latest research papers.
Because we only support C++ there's no overhead or compromise to accommodate other languages. This one is by C++ Devs, for C++ Devs. We have heard it all before, why are you using C++? because it's all written in C++ , that's why !!!
Quick Start Res-Net
Customize the Solver
Quick Start Res-Net
Define the net geometry and automatically generate the layers.
Quick Start U-Net
Customize the Solver
Customize the Solver
Auto generate U-Net architectures
Customize the Solver
Customize the Solver
Customize the Solver
Tune the network by customizing the solver. Choose from base SGD or Adam types solvers.
Pricing Model
Automatic Augmentation
Automatic Augmentation
Using Delve++ to design / train / evaluate networks is just $95.
Email us for commercial deployment of trained networks using the stand-alone evaluation library. Prices from $895 per system, depending of support level requested. We can help with the whole project if desired.
Automatic Augmentation
Automatic Augmentation
Automatic Augmentation
Specify augmentation parameters, to automatically generate multiple images from a single labeled one. Adjust illumination, Noise, Blur, Intensity Histogram, Color, Flips, Rotations.