Deep Learning Lab (Platinum P1)

The Platinum option (P1) of The University program for Deep Learning Lab introduces the concept of Deep Learning accelerated by the GPU. This includes working with popular Deep Learning software frameworks and training convolution neural networks (CNN) in each framework to classify images. The lab also focuses on how to leverage Deep Neural Networks (DNN) within the deep learning workflow to solve real world classification problems using NVIDIA DIGITS. The lab resources also enable the learner to walk through the process of data preparation, model definition, model training and trouble shooting. The Lab also demonstrates the process of validation of data to test and try different strategies for improving model performance using he GPUs. On completion of this lab, the learner will be able to use DIGITS to train the DNN for the customized deep learning application defined by the user. The online course/Tutorial on experimentation helps the users to easily get started working with Deep Learning activities.
This option offers a high computing solution and can be used to introduce the concept of Deep Learning and its frameworks to the learners along with research activities and laboratory experimentation.
  • Lab contains Edutech’s Platinum category of AI/DL machine consisting of server chassis with 1600W redundant power supply, Two 12 core CPU, 128 GB DDR4 ECC RAM, Four GPU each with 16 GB memory and 2500 plus CUDA cores, Two units of 1Tb HDD and 512 GB SSD for fast booting.
  • The software libraries pre-installed should consists of Tensor Flow,Caffe/ Caffe2, PyTorch, CNTK, Theano, DIGITS, Misc: Numpy, Scikit, pandas, other relevant py libs, Essentials: CUDA, cuDNN, TensorRT, OS: Ubuntu 16.04 or 18.04 etc.
  • Datasets like ImageNet, CIFAR-10, KITTI should be pre-loaded for out-of-box development
  • 10 users online learning course/Tutorial to support the Deep Learning related experiments
  • 2 days training for maximum 5 faculties during supply of the setup
  • Soft copy of documentation for experimentation
  • 1 year return to base warranty
Add-on Accessories
In addition to the DL machine, it is suggested to include the following accessories which are required for inferring the trained model on the portable Embedded GPU platform. This platform can be best used for research and project activities and adds a portability feature while implementing the solution.
The following accessories are suggested along with the lab.
  • Embedded GPU board to demonstrate the remote inference of deep learning model
    USB HDD for local host machine booting
  • Portable SSD with  pre-configured tools and OS along with GPU board interfacing cable
  • USB Hub for connecting multiple USB devices like keyboard, mouse, camera etc. to the Embedded GPU board.
Lab Experimentation
Following experiments can be performed using this lab setup:
  • Introduction to Deep Learning
  • Image Classification with DIGITS
  • Object Detection with DIGITS
  • Object Detection over KITTI dataset with DIGITS
  • Semantic Segmentation using DIGITS
  • Medical Image Segmentation using DIGITS
  • Signal Processing using DIGITS
  • Train a Generative Adversarial Network using DIGITS
  • Training an image auto encoder with DIGITS
  • Binary Segmentation using DIGITS
  • Linear Classification with Tensor Flow
  • Image Classification using Tensor Flow
  • Demonstration of remote inference of Deep Learning model using Embedded GPU board